You are on page 1of 50

Number Sense as Situated Knowing in a Conceptual Domain

Author(s): James G. Greeno


Source: Journal for Research in Mathematics Education, Vol. 22, No. 3 (May, 1991), pp.
170-218
Published by: National Council of Teachers of Mathematics
Stable URL: https://www.jstor.org/stable/749074
Accessed: 05-05-2020 22:22 UTC

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms

National Council of Teachers of Mathematics is collaborating with JSTOR to digitize,


preserve and extend access to Journal for Research in Mathematics Education

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
Journal for Research in Mathematics Education
1991. Vol. 22, No. 3, 170-218

NUMBER SENSE AS SITUATED KNOWING IN A


CONCEPTUAL DOMAIN

JAMES G. GREENO, Stanford University and the Institute for


Research on Learning

This paper presents a preliminary attempt to characterize number sense theoretically a


of cognitive expertise. I propose a way to view conceptual domains using a metaph
environment in which one can know how to find resources and use them to make things.
the domain of numbers and quantities as an example of a conceptual environment, and I in
number sense as a set of capabilities for constructing and reasoning within mental mod
perspective provides reasons that support considering various aspects of number
features of students' general condition of knowing in the domain of numbers and qu
rather than skills that should be given specific instruction. Some current trends in researc
cognition and learning support this view of knowing in conceptual domains, including
sense.

The term number sense refers to several important but elusiv


including flexible mental computation, numerical estimation, a
judgment. This paper presents a view of number sense as an exam
in a conceptual domain, the domain of numbers and quantities. I dis
in which conceptual domains are considered as environments in w
know how to live, and people's learning to live in an environmen
from their activities in it. In the view that I present, number sense
cognitive expertise-knowledge that results from extensive acti
through which people learn to interact successfully with the various
domain, including knowing what resources the environment offe
to find resources and use them in their activities, perceiving an
subtle patterns, solving ordinary problems routinely, and generatin

Examples of Number Sense


I believe that number sense is a term that requires theoretical anal
a definition. We recognize examples of number sense, even tho
satisfactory definition that distinguishes its features. We need a theory
important properties of these examples and to explain how those pr
to produce the phenomenon.

This paper was supported by the National Science Foundation, Grant BNS-8
the Institute for Research on Learning. I am grateful to the participants in a
organized by Judith Sowder at San Diego State University in February 1989 fo
discussions that led to this paper. An earlier version appeared in a report of th
edited by J. R. Sowder and B. P. Schappelle. I am grateful to Joyce Moore
Magdalene Lampert, Judith Sowder, John Seely Brown, Susan Stucky, Ro
Resnick, and Robert Glaser for continuing conversations in which I learn about th
I thank Magdalene Lampert, Judith Sowder, Robbie Case, and Paul Dugui
comments on earlier drafts.

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
171

Flexible numerical computation. One form of number sense involves recognition


of equivalences in order to regroup numbers in mental multiplication. An example
was given by Hope and Sherrill (1987) in results of a study of mental multiplication
by high school students. Some students solved the problem 25 x 48 by converting
the problem to o%/4 x 48, then to 100 x 48, then to 100 x 12, to obtain the answer,
1200. In contrast, one of the students solved 25 x 480 saying, "Let's see. 480 on the
top and 25 on the bottom. 5 times 0, 5 times 8 is 40, carry 4, and 4 is 24. I have to
realize that the second number is one over. 2 times 0, 2 times 8 is 16, carry 1; 2 times
4 is 8 and I is 9, so 960, 9600. So 9600 and 2400 is 0, 0,..., 19 thousand
and... 860" (p. 101). The second answer reflects a lower level of number sense than
the first because it operates on a mental analogue of paper-and-pencil symbols,
rather than transforming the problem on the basis of useful equivalences.
Examples of flexible computation involving quantities also have been provided
in studies of everyday reasoning. Scribner (1984) observed young men whose job
was to assemble orders of dairy products for delivery. Orders were written in a
special notation. For each product and container size (e.g., quarts of chocolate milk)
there is a positive integer and a signed integer, for example, 2, -5. The first integer
indicates a number of cases, and the signed integer indicates an additional number
of containers if it is positive or a number of containers less than the number of cases
if it is negative. There are different numbers of containers per case, depending on
the size of the container (4 gallons, 9 half-gallons, 16 quarts, 32 pints, or 48 half-
pints). For example, "2, -5" for quarts of some product would require one full case
of 16 quarts and another case with 11 quarts. Scribner's observations focused on
situations where using an available partial case would allow an order to be filled
more efficiently than using only full cases. For example, if the order is "2, -5" for
quarts and there is a case with nine quart containers, then it is easier to add two quarts
to that case than to take a full case and remove five from it. Scribner used the term
literal to describe a solution in which a completely full or a completely empty case
is used, with the designated number of cases either taken away or added. Sometimes,
of course, a literal solution is optimal. In 53 situations that were observed in the work
setting, literal solutions were optimal in 25 situations, and the workers used literal
solutions in all of them. In 28 situations where an available nonliteral solution was
optimal, the worker used that solution in 23 of the situations.
In another example, Carraher , Carraher, and Schliemann (1985) interviewed
young business persons who sold produce in street markets in Recife, Brazil. The
following exchange occurred with a 12-year-old salesperson, called M:
Customer: How much is one coconut?
M: 35.
Customer: I'd like ten. How much is that?

M: (Pause) Three will be 105: with three more, that will be 210. (Pause) I need four more.
That is...(pause) 315...I think it is 350.
(p. 23)

The salesperson might simply have added a 0 to 35 and read the result, a method
that is especially convenient with written numerals, but not when numbers are

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
172 Number Sense as Situated Knowing in a Conceptual Domain

represented orally. It might be that the price of three coconuts was a landmark in M's
quantity space, and therefore available for use in finding an unfamiliar answer; or
M might have recognized the price of three coconuts as being near 100, allowing
separation of the problem into additions of 100 and additions of 5 for sets of three
coconuts. In either case, the example reminds us that the way that number sense is
exhibited may differ from our expectations.

Numerical estimation. Another kind of performance that is taken as evidence for


number sense involves recognition of approximate numerical values in the context
of computation. For example, given the problem

347 x 6

43

one ninth-grade student said, "It would be easiest to divide the 6 and 43 first, which
is about 7; so 34/7 is about 50" (Reys, Rybolt, Bestgen, & Wyatt, 1982). An example
of students using computational techniques (incorrectly) instead of attending to
magnitudes of numbers, noted by Reys et al. (1982), is in the results of an exercise
in the National Assessment of Educational Progress (NAEP), where students were
asked to estimate the answer to

12 7

13 8

with the instruction, "You will not have time to solve the proble
pencil." Fifty-five percent of 13-year-old students and 36% of 17-
answered either 19 or 21, rather than 2, which was chosen by 24
olds and 37% of the 17-year-olds.

Quantitative judgment and inference. Another aspect of reaso


viewed as evidence of number sense involves judging and making
quantities with numerical values. Back-of-the-envelope problem
1983; Morrison, 1963) provide interesting examples. A study b
included the following example: "How many leaves fall in North
autumn?" One problem solver inferred the approximate area of
from an estimate of the size of Michigan, which she was familia
that Michigan is about average in area. She doubled that result t
parts of Canada and Mexico. She then used a fraction of the tot
covered by trees, adjusting downward for bodies of water and d
inferred a number of trees by recalling a typical distance between
area that she was familiar with, calculating the number of trees pe
that distance, and multiplying by her estimate of tree-covered are
number of leaves per tree from estimating a number of branch
number of leaves per branch and finally inferred a total numbe
product of the leaves per tree and the number of trees.
Anegative example from NAEP, noted by Schoenfeld (1988), in
with 1128 soldiers to be transported in buses that hold 36 soldie

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
James G. Greeno 173

asked how many buses w


of students was "31 re
operation without a se
without regard for the

Number Sense and Instruction

Several commentators have agreed that improved number sense is a desired


outcome of mathematics education (e.g., National Council of Teachers of Math-
ematics [NCTM], 1989; National Research Council, 1989; Blackwell & Henkin,
1989). A natural response is to treat flexible mental computation, computational
estimation, quantitative judgment and inference, and other indicators of number
sense as skills and to try to design instructional activities in which students can
acquire the skill. An alternative approach would treat the various manifestations of
number sense as symptoms of a more basic and general condition of knowing in the
conceptual domain of numbers and quantities. In this more global treatment of the
question of number sense, the growth of students' number sense could be consid-
ered as resulting from the whole range of activities of mathematics education, rather
than a designated subset of specially designed activities.
The global view of number sense has been expressed well by Howden (1989):

Number sense can be described as good intuition about numbers and their relationships.
It develops gradually as a result of exploring numbers, visualizing them in a variety of
contexts, and relating them in ways that are not limited by traditional algorithms. Since
textbooks are limited to paper-and-pencil orientation, they can only suggest ideas to be
investigated, they cannot replace the "doing of mathematics" that is essential for the
development of number sense. No substitute exists for a skillful teacher and an
environment that fosters curiosity and exploration at all grade levels. (p. 11)

Although the activities of learning to recognize patterns and equivalences may


include some deliberate pattern-finding and descriptions, many of the properties
and relations involved may be largely implicit, involving holistic and configural
understanding rather than deliberate rule-based procedures. Indeed, it may be more
fruitful to view number sense as a by-product of other learning than as a goal of
direct instruction. In Cobb and Merkel's (1989) terms,

The activities were not designed to lead pupils to "see" specific relationships.... Instead,
their function was to give the children opportunities to think about what they were doing
as they solved arithmetic problems. (p. 72)

The "environment that fosters curiosity and exploration" is a social construction


in which students interact with the teacher and with each other about quantities and
numbers. In such environments, teachers and students are engaged in conversations
where they develop and negotiate meanings of terms and make sense of numbers
and quantities in situations, as Markovitz (1989), Reys (1989), and Silver (1989) all
pointed out. They also attend explicitly to processes of thinking, as Carpenter (1989)
and Lampert (1990) have emphasized. Examples of conversations in the domain of
numbers and quantities are found in classrooms that are organized around the
activities of collaborative learning and understanding. These include primary

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
174 Number Sense as Situated Knowing in a Conceptual Domain

school classes such as those based on problem solving (Cobb & Merkel, 1989) and
construction and discussion of methods of solving word problems (Fennema,
Carpenter, Keith, & Jenkins, 1989). Lampert's (1986; 1990) teaching of mathemat-
ics in fifth grade is designed as a collaborative activity in which she and the students
work together to understand mathematical concepts, notations, and procedures. At
the Farm School in Irvine, California, students spend a significant amount of time
in games involving numbers, including a continuing game where one student
provides a few examples of numbers related by a function, and other students then
construct further examples based on their hypotheses about the function (Lave,
Smith, & Butler, 1988).
Practice in estimation and quantitative judgment occurs naturally in conversa-
tions about quantities involving different levels of resolution, as when the group
deals with imprecise aspects of everyday problem solving (cf. Hope, 1989). If one
is talking about the amount of food that a whale consumes or the amount of money
that is needed to purchase supplies for a trip, it is clear that exact answers are
unnecessary for most purposes. Therefore, socially organized activities that include
discussions of quantities may be more successful for students' learning to reason at
appropriate levels of detail than exercises with directions to find approximate
answers.

Activities that explore relations among numbers and quan


construction of patterns and the identification of equivalen
Viewing mathematics as the science of patterns (Steen, 1
technology of procedures, would be consistent with this emphas
the patterns and equivalences have utility, and finding altern
answers can contribute to students' knowing the environme
quantities as well as to their understanding of how mathem
situations involving physical objects, quantities of money, a
things.

KNOWING IN CONCEPTUAL ENVIRONMENTS

In this section I discuss some general ideas about conceptual domains v


environments. Some readers may prefer to read the next section first, wher
the domain of numbers and quantities specifically, and apply the ideas
hypothetical explanations of examples of number sense.
A subject-matter domain is a structure of facts, concepts, principles, proc
and phenomena that provides resources for the cognitive activities of k
understanding, and reasoning. We usually think of knowledge of the do
subset of that structure in a person's mind. In the current information-p
framework of cognitive science, knowledge is a set of representations that a
in the mind, including symbols that represent concepts, properties, and rel
well as representations of procedures for manipulating symbolic expre
Learning a domain, in this information-processing framework, is the con
of cognitive structures and procedures that represent the concepts, princ
rules of inference of the domain.

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
James G. Greeno 175

Knowing and Learning


I propose an alternativ
thought of as an environ
metaphor, knowing the
knowing how to use it
available in the environ
for understanding and
ment in its own terms
understanding how its
includes knowing what r
your individual and socia
resources productively.
to live in an environme
available, and learning
productively and enjoya
Like the information-
involves access to conce
enriching the known set
way that known concept
In the information-pro
cognitive structure, and
sentations and interpre
knowing a set of conce
concepts but rather inv
processes of reasoning.
important role in reason
resources in a physical e
her ability to find and u
instructions as the basis
People inhabit many dif
natural environments-
beauty and sustenance,
biologicaland geological
large human-made envi
resources for buying g
interesting architecture
ments with materials an
environments are simil
that occur naturally in
deliberately constructed
munities in academic d
conceptual environment
Living in an environme
other people and collabo

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
176 Number Sense as Situated Knowing in a Conceptual Domain

environment for the benefit of families, groups of friends, and social and business
organizations, as well as for one's individual benefit. People learn to live in
environments through a combination of social and individual activities. A person
can learn his or her way around in a city by travelling with a friend who already
knows how to find something, by exploring parts of the environment with other
newcomers, or by following instructions given by a friend for finding something
and thereby learning the meanings of symbols such as the names of streets or
buildings. Learning also occurs in individual activities, often using maps or guide
books that have been constructed in the social community using socially accepted
conventions of representation and that then can be used by individuals as guides for
their individual problem solving and exploration.
Learning to use resources to make things also involves social as well as individual
activities. In a kitchen or workshop, a person can learn how to put materials together
and perform operations on them (e.g., various mixing operations, different methods
of cooking, ways to attach parts, and how make adjustments and repairs) by helping
a more experienced worker, by being shown how to do things and then trying them
with coaching by a mentor, or by following written instructions and observing the
results.

Learning a conceptual environment also involves a combination of social and


individual activities. Written texts, diagrams, and computer programs can provide
representations of the resources that a conceptual domain has, and an individual can
use these representations to guide his or her exploration of the conceptual environ-
ment. Learning a conceptual environment by oneself, however, is much harder than
learning it with others, especially with the help of others who already know their
ways around and how to use the resources of the domain in making inferences.
Furthermore, individuals who learn a conceptual environment on their own are
likely to be less successful in the social aspects of intellectual work in the domain.
A person's knowing of a conceptual domain is a set of abilities to understand,
reason, and participate in discourse. As Hesse and Anderson (in press); Posner,
Strike, Hewson, and Gertzog (1982); and Toulmin (1972) have discussed, any
particular activities that a learner engages in or learns to perform are embedded in
a conceptual ecology that has been developed within a community of intellectual
work, such as the members of an academic discipline. Critical components of these
sets of practice include the appreciation and use of explanatory ideals that are shared
within the community and provide basic modes and goals of explanatory discourse.
It is possible to learn about an environment without living there, by studying maps
and descriptions of the place and the activities of its people. Such learning, however,
is largely limited to interactions with symbolic expressions-speaking or writing or
drawing maps-which are very different from the abilities needed to live and work
in the environment successfully. In learning a conceptual domain, it is possible to
confuse representations of concepts with the concepts themselves and to learn how
to manipulate symbolic expressions rather than how to find and use conceptual
resources. In too many cases, classroom instruction in mathematics and other
conceptual domains is analogous to learning about an environment by studying

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
James G. Greeno 177

guidebooks and maps


recipes without ever
Howden (1989), that
numbers, visualizing
should view symbol
written or spoken ins
of learning to inhabit
and communication, b
ments as the main lea

Hypotheses About R

The environmental
concepts are in a doma
and how to move abo
about what people do
presents ideas about a
of an environment su
to make things with
proposal, th I suggest
constructing and ma
entities and recognizi

Mental models. I hyp


a major use that peop
of situations that the
These situations are co
to objects in the situa
A mental model is a s
and behavior of symb
of the objects they re
mental model differs
or formulas that expr
the person interacts w
and manipulating thos
objects or people in a
model are based on th
people, called enactm
expressions of propos
Reasoning with ment
very preliminary sta
analyses by Fauconni
a volume edited by G
mental model include
simulating a walk thr
in each of the rooms,

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
178 Number Sense as Situated Knowing in a Conceptual Domain

to decide which to take (Chase, 1983). More complex examples are thought
experiments in physics, such as those described by Einstein (1961) in clarifying
ideas about relative motion.

Operations with some of the objects in mental models are like operations with
physical objects. In reasoning about these objects, the person mentally moves them
about, moves about on or in them, combines them, separates them, changes their
sizes and shapes, and performs other operations like those that can be performed on
objects in the physical world. Operations with some objects in mental models are
like operations on complex physical systems, such as pulling on a rope that is part
of a pulley system (cf. Hegerty, Just, & Morrison, 1988), turning the steering wheel
of a car, or moving an icon on a computer screen. Operations with some objects in
mental models are like things we do with other people. In reasoning with mental
objects that represent people, the person mentally performs some of the activities
of interpersonal interaction, such as contributions to a conversation, facial expres-
sions, and gestures.
A mental model works because operations with mental objects in the model have
effects that are like the effects of that operation on the objects that the model
represents. The behavior of objects in the model is similar to the behavior of objects
that they represent, and inferences are based on observing the effects of the
operations. Abilities to reason with mental models depend on a kind of knowing in
which the objects in mental models behave according to constraints that have
productive consequences. These constraints correspond to principles that operate in
the domain of objects that the model represents, but the principles are implicit in the
behavior of the objects, rather than being stated explicitly. According to this
hypothesis, learning to reason with models that have appropriate constraints is a
crucial part of learning the domain.
An example of a geometric mental model was discussed by Simon (1978).
Imagine a rectangle one unit high and two units wide. Now imagine a vertical line
that divides the rectangle into equal parts. Next, imagine a straight line from the
upper left corner to the lower right corner of the original rectangle. Finally, imagine
a straight line from the upper right corner of the original rectangle to the point where
the vertical line in the middle meets the bottom of the rectangle. Can you answer the
following questions: (a) Do the two lines that were drawn diagonally intersect? (b)
On the last line that was drawn, is the part above and to the right of that intersection
longer or shorter than the part below and to the left of the intersection?
The ability of most people to answer questions (a) and (b) correctly illustrates a
widespread ability to construct and reason with mental models. The process works
because the representations of lines in the model simulate properties of lines in a
plane. In particular, (1) when there is a line in a plane and another line is drawn from
one side to the other of the first line, the lines intersect, and (2) when a diagonal of
a square from lower left to upper right intersects another line that passes through a
point somewhere on the left vertical side and the lower right corner, the intersection
of these lines on the diagonal is nearer the lower left corner of the square than the
upper right corner. The answers to questions (a) and (b) could be inferred if there

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
James G. Greeno 179

were a set of expression


is not how it is done w
properties that are st
properties of the obje
antecedent and consequ
that the objects in the
Answers to questions
of paper or constructi
diagram. I propose tha
there are representation
lines in a plane and the
in the r representation,
of conditional proposi
into the behavior of a
of inferences that are
1978. constructed com
some operations on di
The idea that people i
interact with objects (ra
is consistent with an o
physicists and physics
made as they worked.
like those the individ
example involved a prob
working on the problem
size of a grapefruit an
direction that the earth
he waved his left hand
sunrise; then with his l
right hand in a counte
while he referred to th
"is going in the same d
He then said "Is that ri
right, ok." It seems lik
that represented the e
represented the ship. O
moving the "ship" acr
simulated motions was
to the problem. The ph
operate on his mental
relative movements of
of reasoning with the
with a model that has
results of the movements.

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
180 Number Sense as Situated Knowing in a Conceptual Domain

Representing concepts in models and recognizing conceptual structure. Model-


based reasoning depends crucially on the objects that the model contains and the
constraints that govern the behavior of those objects. I propose that a major part of
knowing a conceptual domain is an ability to construct mental models that conform
to the concepts and principles of the domain.
Construction of a model with conceptual entities requires a kind of knowing of
the domain that has those concepts in it. I suggest that knowing how to construct
models in a domain is like knowing how to work in an environment that has
resources for a kind of constructive activity, such as a woodworking shop or a
kitchen.

A shop or a kitchen has objects, materials, and tools that can be used to make
things. Knowing how to work in such an environment includes knowing what
objects and materials are needed for various constructive activities, knowing where
to find those objects and materials in the environment, knowing what implements
and processes are useful for constructing various things, knowing how to find the
implements, and knowing how to use the implements and operate the processes in
making the things that can be made.
In constructing conceptual models, the ingredients are representations of specific
examples of concepts. For example, to make a graphical model of a problem
involving functions, one draws some lines, either on paper or in a mental model. To
construct a force diagram for a physics problems, one draws some arrows. We can
think of the conceptual domain as an environment that has representations of
concept-examples stored in various places. Knowing where to find these, knowing
how to combine them into patterns that form models, and knowing how to operate
on the patterns constitute knowledge of the conceptual domain. The representations
of concept-examples have to be understood in a special way. They are not only
objects that are drawn on paper or represented in the mind. They are objects in the
stronger sense that their properties and relations interact in ways that are consistent
with the constraints of the domain.
When someone knows concepts and principles of a domain, those concepts and
principles are represented either implicitly or explicitly in the mental models that
he or she constructs and with which he or she simulates events. For concepts that
are represented explicitly, the person's knowledge is like knowing where to find the
concepts and put examples of them into mental models. When a concept or principle
is represented implicitly, the person knows how something in the environment
works, and objects in the model work that way.
Consider an example of implicit representation from physics. An interviewer
shows a picture of an airplane flying and asks a participant to draw the path of an
object that is dropped from the plane. Many people draw a straight vertical line, and
this is sometimes interpreted as evidence that people have an implicit theory in
which impetus is imparted to objects when they are propelled (e.g., McCloskey,
1983). The answer usually considered correct is to draw a parabolic line, consistent
with the Newtonian principle that an object in motion tends to remain in motion in
the same direction and that a force such as gravity changes the velocity as the object
moves through space.

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
James G. Greeno 181

The answers that many


models in which they ena
et al., 1988). Someone w
airplane and someone el
have objects in their me
physicists can describe
involving different conc
may not have explicit re
would be possible to gene
an explicit representatio
Another example inclu
Roschelle and Greeno (19
sional physicist who was
asked "What is happenin
blocks moving and of for
inferred that the phys
forces, probably in the fo
are used to represent ma
As a mathematical exam
equation such as v = 6 -
function, its v-intercept
equation, %? + / = 1, rep
other properties: the x-in
represented by a straight
Graphs are models in w
of lines are useful in mak
the function that v = 6
negative. That property c
not enter the lower left
3x and v = 3x have a solu
because the lines interse
the equations. A LISP pr
that y= 6 - 3x if
and = 3
solutiony= 3, x= 1. The p
properties of equations a
equations are used, the p
symbols, how many cha
cantly to properties of th
a graphical model, the lin
relations that correspond
A graph drawn on paper
we think of mod mental
object, at least entirely
model, rather than a phy

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
182 Number Sense as Situated Knowing in a Conceptual Domain

solvers to imagine an object in motion to solve a problem about relative motion. In


many cases, a combination of physical and mental modelling occurs, as when a
diagram is drawn for a physics problem and the person reasons about properties of
objects depicted in the diagram if they are set in motion (Roschelle & Greeno, 1987),
or when graphs of two functions are put on a piece of paper and the person reasons
about the effect of translating one of them.
Another function of knowing concepts in a domain involves recognition of
properties and relations in situations that are experienced. Concepts enable people
to perceive things in different ways, allowing different ways of "seeing-as,'" to use
Hanson's (1961) term. One of Hanson's examples involves concepts of astronomy.
One can see the sun as a disk that travels across the sky. One also can see the sun
as a very large, very distant body that is visible part of the time because the planet
we are on constantly rotates. Another example involving physics is the recognition
of forces, including friction, when one is sliding a heavy object across a floor.
Similarly, one can recognize an additive or multiplicative mathematical function
relating two variables in a physical or commercial situation.
Mental models can be used to arrive at different ways of understanding problems
and situations. This was illustrated by an example given by Clement (1989), who
asked professors and graduate students in technical fields to solve the following
problem:

A weight is hung on a spring. The original spring is replaced with a spring-


"* made of the same kind of wire.
"* with the same number of coils.
"* but with coils that are twice as wide in diameter.

Will the spring stretch from its natural length more, less, or the same amount under
the same weights? Assume the mass of the spring is negligible compared to the mass
of the weight). Why do you think so?
(p. 350)

In one solution, the problem solver constructed models of hypothetical modifi-


cations of the spring, including a zigzag in which stretching the spring corresponded
to increasing the angles where the wire bent. A satisfactory solution was obtained
when the problem solver thought of a spring as a wire bent into a square with a
weight pulling down at one of the corners. In this model, stretching the spring
corresponded to twisting the wire. This caused the problem solver to see the amount
of stretching as dependent on torsion forces and to conclude that a longer wire would
bend more with the same amount of force.
If this idea about conceptual understanding and reasoning is approximately
correct, then an important question for mathematics education is how students learn
to construct and reason with mental models that have objects representing math-
ematical concepts. The importance in mathematical learning of forming conceptual
entities or objects of thought has been recognized, for example, by Ayers, Davis,
Dubinsky, and Lewin (1988), by Beth and Piaget (1966), by Harel and Kaput (in
press), and by Kaput (1987).

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
James G. Greeno 183

Conceptual Models in
In a community that u
conceptual entities of t
problems, tell about in
discovered, ask whethe
and so on. These are an
environment, such as
hardware store where
house that avoids traff
that people have about
ingredients or about w
available kind of finish
A condition for any co
the terms they use ref
the conversation is abou
some problems of refe
participants have in min
to focus the participants
the same modes of cogn
has called alignment, w
Achieving shared refere
complicated. Scheglof
about such as locations,
participants reached st
mind. In conversations
ticipants also need to de
things they are discuss
I suggest, as a hypoth
different from the one
simulate properties of t
(Fauconnier, 1985, and
language understanding
communicate about loc
environment, and I hyp
construct a mental m
described. It seems easi
landmarks, and this m
easier when some of t
places sometimes fail
scientific meeting to me
the convention registrat
were having the conver
consistent with the te
consistent with each other.

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
184 Number Sense as Situated Knowing in a Conceptual Domain

In conversations about conceptual environments, I hypothesize that people use


mental models in the same way that they use them in conversations about physical
environments they are not in. Such a conversation succeeds if the participants
construct mental models that coincide in the ways that are required for terms to have
the same referents, and if the terms are interpreted as referring to the same objects
and properties in the two models. For example, in a mathematical conversation,
someone may refer to two linear functions that intersect at the point (0, 4). For
understanding to occur, the speaker and listener would both construct mental
models with two objects corresponding to the functions, an object corresponding to
the point (0, 4), and relations in which the representatives of the functions both pass
through the representative of the point. The objects in these mental models have
properties that were not specified in the initial description but that can be mentioned
or assumed in further conversation. For example, the function with higher values to
the right of (0, 4) has lower values to the left of (0, 4), and neither function goes
through the point (0, 0). Then, for example, one of the people could then refer to "a
function that goes through the origin," and this would be understood as being
different from either of the functions that were already mentioned.
The degree of specificity needed in such a model is an open theoretical question,
in my opinion. Someone might imagine two lines going through the point (0, 4), or
there might be a partially specified mental representation in which the slopes of the
lines are not determined. I would expect that the specificity would be sufficient to
support some inferences. For example, the fact that whichever function has lower
values of y for negative x has higher values of y for positive x would be seen easily
in an image of two specific functions; it might also be a property of partially
specified objects that correspond to functions whose slopes are not determined but
have the property of crossing at a designated point.'
Conversational processes are available to disambiguate references to objects in
mental models. As an example, if one person referred to two functions that go
through (0, 4), the other person could ask, "Are they both linear?" or "Are their
slopes positive?" or other questions that would provide information that seemed
necessary to proceed with the discussion. Differences between the models that the
different participants construct can lead to difficulties in understanding, and
conversational repair can become necessary for such cases. For example, one of the
individuals might refer to a function that goes through (0, 0) and the intersections
of that with the first two functions in the upper-right quadrant. If the other person
has a model in which the first two functions have positive slopes, he or she might
object that a function through (0, 0) might not intersect both of the first functions-
or either of them. It might turn out, then, that the first person's model had functions
with negative slopes and the function going through (0, 0) was positive. Conver-
sational repairs in this case would deal with properties of the functions that had to

'This idea differs from a classical assumption about mental images, associated with Berkeley (e.g., Jones, 1952), that
images have to be specific in every respect. Berkeley asserted that it is impossible, for example, to imagine a triangle
without its being a specific triangle - it would have to be equilateral or isosceles or scalene, but not indeterminate. I
assume that mental models can include objects with some indeterminate properties.

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
James G. Greeno 185

be agreed on for the peop


It is likely, in this kind
support their conversatio
situation and permit str
pointing to the lines in t
Although shared refere
conceptual entities them
facilitate communicatio
about a function in a wa
about the location of a ho
we rely especially on re
students' learning may b
interact with the symbol
conceptual entities that
symbolic representations
for example, reasoning a
supported quite well wi
represent graphs. Other
understanding and rea
constructing a mental mo
in which elements of th

THE CONCEPTUAL ENVIRONMENT OF NUMBERS AND QUANTITIES

The metaphor of a subject-matter domain as a conceptual environment, with


knowing the domain as the ability to find and use concepts and principles in the
environment as resources, seems to apply well to the domain of numbers and
quantities, particularly in its suggestions regarding number sense. Numbers and
quantities are important objects in the domain with a structure of relations and
operations. People with number sense know where they are in the environment,
which things are nearby, which things are easy to reach from where they are, and
how routes can be combined flexibly to reach other places efficiently. They also
know how to transform the things in the environment to form other things by
combinations, separations, and other operations. As a hypothesis, I propose that an
important function of knowing the lay of the land is enabling individuals and groups
to construct mental models that contain conceptual entities and to reason with those
models by operating on the mental objects that represent numbers and quantities.
The metaphor of an environment seems to work for the domain of numbers at two
levels. At one level, the domain is a large and complex environment, constructed by
human workers over many hundreds of years, with important general features,
comparable to commercial and industrial districts and residential neighborhoods.
Mathematical features at this level include different kinds of numbers, such as
integers, rationals, and reals, and different quantitative domains, such as commer-
cial transactions, cooking, and motions of objects. General mathematical features
also include operations on numbers-addition, subtraction, multiplication, division,

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
186 Number Sense as Situated Knowing in a Conceptual Domain

powers, roots-and operations on quantities-additive or multiplicative combi-


nations of quantities, processes of exponential growth, and so on. Beyond numbers
and quantities, of course, the mathematical environment includes sets, functions,
variables, derivatives and integrals, and many other concepts that are constituted by
different kinds of mathematical objects and properties.
In the domain of quantities and numbers, a quantity is an amount of something.
Any physical object or event has many quantitative properties. Each quantitative
property can be described numerically by assigning a unit of measure. Numbers,
then, are included in descriptions of quantitative properties of objects. Given a
choice of units, a number is a property of an object or event. Numbers are also
conceptual objects in their own right. They can be considered either as elements of
the system of numbers or as reifications of properties. In reasoning about quantities,
both of these modes of cognition occur. For example, in the calculations for a
problem, numerical operations are carried out that apply to the numbers regardless
of their quantitative embeddings.
Someone who knows the environment of numbers and quantities knows how to
find the kinds of conceptual objects that each region of the landscape contains, and
knows relations among them. These relations include correspondence between
operations on numbers and operations on quantities, such as multiplicative relations
that correspond to systems in which quantities are related by proportions. They also
include operations that can be applied to different kinds of numbers, and the
similarities as well as the differences in operations that are applied to numbers of
different kinds, such as those between addition and subtraction of integers and of
fractions.

The metaphor of an environment also applies locally, at the level of a workshop


or kitchen. The mathematical objects used for solving specific problems include
representations of numbers, with operations on the representations that simulate
operations on numbers, such as addition by combining the representations of two
numbers or multiplication by expanding the representation of one number by a
factor equal to the other. Knowing the environment of the mathematical workshop
or kitchen includes knowing the locations of the various numbers to be represented,
knowing which of them are near each other, knowing how to combine representa-
tions of various numbers to form representations of other numbers, and knowing
various ways to make representations of different numbers by combining, separat-
ing, expanding, and contracting other representations.

Mental Models of Numbers and Quantities

When someone solves a problem involving numbers or quantities, different


mental processes can be used. I will focus here on one kind of process that involves
construction and reasoning with a mental model.
When a problem is presented in symbolic form, a problem solver interprets the
symbols in some way. The interpretation gives meaning to the symbols and provides
a way or some ways of reasoning to solve the problem. Aphysical or mental model
provides one way to interpret symbolic expressions. When a model is used, physical

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
James G. Greeno 187

or mental objects are pla


or presupposed by the e
in the expressions. The
combining, separating,
operations can provide i
of a problem.

Numerical objects and operations. First, consider problems of numerical calcu-


lation, such as 25 x 48, (347 x 6)/43, or 2/13 + 7/8. One way of interpreting such
problems involves constructing models with objects that correspond to numbers.
The model could be physical, with objects such as piles of blocks or pictures on a
computer display, or the model could be mental with mental objects that represent
the numbers. To be useful, the objects in the representation should have properties
that simulate properties of numbers, and effects of operations on the objects should
simulate the effects of operations on the numbers that the objects represent.
For problems that are related to number sense, the relevant properties involve the
magnitudes of numbers. Objects that have properties of magnitude such as length,
height, area, or volume can correspond to numbers in mental models. If signed
numbers are included in problems, objects in models need both magnitude and
direction, such as columns above and below a line that is taken to be zero.
To serve as a model for calculation, operations on objects need to simulate
operations on numbers. If the sizes of objects correspond to numbers, then
operations that do this include combining two objects, corresponding to addition,
or removing part of an object, corresponding to subtraction. Positions and distances
along an object also can correspond to numbers, for example, when a model has a
column that extends above and below a zero line, and addition or subtraction of a
number corresponds to moving upward on the column by a distance that corresponds
to the addend or moving downward on the column by a distance that corresponds
to the minuend.

General operations on physical objects that correspond to multiplication and


division are less obviously available than those that correspond to addition and
subtraction. Operations that involve specific factors or divisors, however, are easy
to construct. Doubling a number can correspond to taking two objects that
correspond to the number and combining them; halving a number can correspond
to taking an object that corresponds to the number and dividing it into two equal
parts. It seems reasonable that people can operate on and judge objects directly for
lengths that are approximately twice or half of each other, that tripling and dividing
in thirds probably are available to some people, and that combinations of two
operations such as multiplying by four via two doubling operations also probably
are available.

I hypothesize that operations in mental models can simulate physical operations


that might require special systems devised for the purpose. For example, physical
devices could be designed to perform physical multiplication and division directly
by any number for which a device was available. One such device would be a five-
multiplier consisting of a rod with two pointers that both move when either of them

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
188 Number Sense as Situated Knowing in a Conceptual Domain

is pulled along the rod, and that move such that one of them is always five times as
far from the end of the rod as the other. Then, multiplying a number by five would
correspond to having an object the length of the number, setting the multiplier next
to the object so that one end of the object coincides with the zero-end of the
multiplier and moving the near pointer to coincide with the other end of the object.
Then the distance from the zero-end to the far pointer of the multiplier corresponds
to the product of the given number and five. The device could be used to physically
divide lengths as well as to physically multiply them. A multiplier like this could be
constructed for any given number.

Locations and relations of numerical concepts. Learning the structure of the


domain of numbers goes on over many years. The structure that children learn
initially is linear order. Numbers are in a line, and children learn to recite their names
in order. They learn to use that sequence to count the number of things in a set-that
is, they learn the quantitative property of cardinality. They also learn that numbers
themselves have magnitudes and that the sequence of numbers is ordered by
increasing magnitude-a number later in the sequence is greater than a number that
is earlier.
The linear structure of positive integers is understood easily by children, and
Gelman and her associates (e.g., Gelman & Gallistel, 1978) have argued that the
principles of order, one-to-one correspondence, and cardinality that underlie early
number understanding are among the constraints that young children bring to the
learning of numbers, rather than having to be learned. This might be because order
and correspondence are direct correlates of spatial and spatio-temporal properties.
This could explain why the ability to construct mental models with numbers as
objects is easy to acquire: the objects can be located according to a simple spatial
principle. It could also explain why the relation of one-to-one correspondence
between numbers and objects in a set should be easy for children to identify,
especially when the objects in the set are arranged in a line, because the correspondence
involves coordination in time between attention to two linearly arranged sets.
Cardinality requires cognizance of a group of objects that can have a numerical
property. The grouping of objects can correspond to a spatial property, and
cardinality probably requires coordination of grouping, order, and one-to-one
correspondence (Greeno, Riley, & Gelman, 1984).
If numbers are initially objects that are arranged linearly, then operations on
numbers can be learned and understood spatially as well. Children's early under-
standing of addition and subtraction is grounded in counting operations (Fuson,
1988; Steffe & Cobb, 1988), along with other spatial properties such as the
comparative sizes of objects (Resnick & Greeno, 1990). Growth of the ability to
reason numerically with mental models occurs gradually, originating in direct
interactions with physical objects, such as touching objects or pointing at them, and
progressing through the ability to reason numerically with objects that are perceived
and the ability to reason numerically with conceptual objects that are not perceptu-
ally present (Steffe & Cobb, 1988).
Abilities to reason numerically grow to include operations of addition and

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
James G. Greeno 189

subtraction considered
in mental models, the
previously. The spatia
coordination of counti
combining and separat
numbers and distances
Mental models of mul
as proportional expansio
as additive operations
Proportional sizes of o
arein children's experie
ing models in which mu
exchanges are common
that relevant experien
computational systems
I hypothesize that c
combinations and deco
with the Trafton's (198
a physical sense of num
"knocking off' parts a
combinations that it ca
posed into. As an exam
to form 9, it can combi
into 1 and 7, or into 2
combine with 2 to for
decomposed into 2 and 4
64, and it can be decom
Some empirical eviden
Kilpatrick, and Cunnin
the similarity of pairs
a number from zero t
names, Arabic or Rom
marks, regular n-sided
of fingers. Participants
the similarity of the n
numerals but ask the p
correspond to the Arab
the symbols represent.
First, judgments depe
rather than on the for
asked to judge the simi
theirjudgments were ve
rows of dots, spoken na
found it natural to co

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
190 Number Sense as Situated Knowing in a Conceptual Domain

similarity of those mental objects, with these judgments influenced negligibly


the physical form in which the stimuli were presented.
A second finding concerns the similarity judgments made for abstract conce
Participants were asked to judge similarity of number concepts correspondin
spoken words, rows of dots, and Arabic numerals; and the judgments were v
similar in all three conditions. Multidimensional scaling was used to represent
data. A two-dimensional solution fit the data well. In this solution, numbers w
approximately ordered in one dimension according to the order of their magnitu
In the second dimension, multiples of two (two, four, and eight) formed one clu
multiples of three (three and nine) formed another cluster (with six between the
clusters), and the primes five and seven were neighbors. I interpret this as evide
that multiplicative relations provide a basis of similarity between the numer
concepts, so that when one number is an even multiple of another, that cause
pair to be judged as being more similar. Shepard et al.'s representation of t
concepts as being located in space provides a direct expression of the idea that
objects in a conceptual domain can be thought of as being located in an environm

Quantitative operations and relations. Now consider problems about quantit


Most word problems are about numerical properties of quantities. Amental m
can have objects that represent quantities, similar to objects that represent num
Measures of quantities differ from numbers, however, in having units and in th
being constraints on operations on the quantities that do not apply to number
Knowing a quantitative domain includes knowing how its various quantiti
interact. For example, quantities such as lengths can be combined by concatenati
and measures of such quantities can be combined by addition, giving the nume
value of the concatenated combination of quantities. For the numerical additio
be valid, however, the numerical values have to be in the same units.
Compositional constraints also apply to quantities involving different dime
sions. A number of units and quantity of price per unit can be combined to form
quantity of total price; a quantity of volume and a quantity of density can
combined to form a quantity of mass; and so on. In many domains involvin
physical processes, rates of quantitative change over time are important quant
and a rate of change in a quantity can be combined with a quantity of time to pro
a quantity of the amount of change.
I hypothesize that knowing the quantitative structure of a domain includ
knowing the compositional relations of quantities in the domain. This means that
person knows conceptual entities corresponding to the quantitative concepts of
domain, and that properties of each of these conceptual entities include ways
it can be combined with other quantities in the domain and ways it can
decomposed into other quantities.
Learning the quantitative structure of a domain probably occurs through a l
amount of reasoning activity involving operations on representations of quantiti
Forbus and Gentner (1983) have sketched a hypothesis in which knowing ab
quantities rests on knowing about qualitative processes in a domain. Some com
tational learning environments present physical models of processes designed

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
James G. Greeno 191

show relations among q


Suir, & Grosslight, 198
velocity and accelerati
displays can be effective
understanding that have t
mental models for reaso

Reasoning with physica


of numbers or quantities
make them useful for r
tant. The potentials for
quantities to form other
to simulate assembling a
the making of a sauce ou
In many cases, both in p
the numerical result of a
numbers or quantities i
According to the view o
model that represent the
interaction of the perso
retrieval of information
I hypothesize that men
conceptual entities that th
with a representative of
potential for being comb
thatcontains representat
conceptual objects by ad
multiplying to form a re
of 2. These operations are
ingredients that can be co
pieces of wood of differe
kinds of objects.
Reasoning with mental m
that are represented. Spa
and connecting paths. Wh
what places are nearby an
that are connected by f
tions also seems to be a p
when a person imagines b
other rooms are and how
that such proximity and
based on conceptual dom
placed in a mental model
entitiesin the conceptua
representative of a num

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
192 Number Sense as Situated Knowing in a Conceptual Domain

what other numbers are near the numbers that are represented in the model, as well
as other numbers that are connected to the numbers represented in the model by
familiar paths. (I suppose that familiar paths consist of associative links that
correspond to familiar operations, such as numbers that can be reached by multiply-
ing by ten or by squaring.)
I hypothesize that proximity and path-connection relations support approximate
reasoning because a mental model can be constructed with representatives of
numbers near or connected to the numbers given in a problem, and that the
compositional properties of these nearby landmark numbers are then available for
reasoning.
Conceptual objects that represent numbers can have many properties, and a
person can therefore attend to different properties, depending on the context. For
example, each object that represents a rational number in a mental model might have
two parts, each of which also represents an integer, and properties of each object
representing an integer correspond to which other numbers are factors of the integer.
A person could focus attention on the denominators of two rational numbers, and
attend to their factors, and perceive whether they have a factor in common.
The possibility of selective attention applies also to objects in mental models that
represent quantities. A quantity has many properties, including its magnitude in
relation to most quantities of its type that it is compared with, its physical units, and
its numerical value. A person can focus attention on any of these and thereby
perceive affordances for different inferences and judgments. (The term affordance
was coined by J. J. Gibson, 1986, to designate how things in a situation contribute
to the possibility or impossibility [more generally, to the ease or difficulty] of
activities that people can engage in.)
Assuming that mental models have properties of spatial environments also
suggests that people can focus attention on properties at different levels of detail.
The visual system has information at different degrees of resolution; general
contours can be perceived over a large area, or more detail can be perceived in a
smaller area. Regarding locations of objects in an environment, regions of a spatial
environment can be treated as objects; a person can know that he or she is in one
region rather than another, such as being in one's office rather than in the room where
one teaches a class, and a person in one region can think about a different region and
decide to go there. At the same time, regions have internal structure. Within a region,
objects have specific locations, and one finds one's way around, arriving at specific
locations in order to perform specific actions, such as sitting in a chair and typing
at a keyboard. In a mental model of numbers or quantities, levels of resolution
enable attention to be focused on general features, such as the approximate relative
magnitude of two quantities, or on details, such as the ratio of the measures of the
quantities, accurate to four significant digits.
The properties of conceptual objects in mental models depend on a person's
knowledge of the domain. If a child only knows numbers as objects used in
counting, the representatives of numbers in his or her mental models have properties
that correspond to their positions in the ordered sequence, but they do not have

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
James G. Greeno 193

propertiesthat correspo
numbers. We could say
domain do not yet have
properties and the child d
these we choose to say.
whether the relevant pr

Conjectures About Num

The following discussio


paper suggests how the id
apply to specific exampl

Flexible computation.
equivalence among object
as Behr (1989) and Traf
same as other objects or
expect of a person who kn
examples at the beginnin
that represent numbers,
as quantities, and a ment
Consider the example
problem 25 x 48. The fact
25 and 4 are factors of
to a hypothesis of ment
model with objects repr
many properties, among
numbers. When represen
multiplicative relations, t
an emergent property th
100 x 12. This emergent
the placement of two lin
property of their interse
The example from Scri
computation in which th
filling an order involved
specified number of cont
case. One possibility is t
interpreted as a number
positions in a case. The w
with this goal and see w
either by increasing or d
specification, or by incre
agree with a- specificatio
as potential states of affa

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
194 Number Sense as Situated Knowing in a Conceptual Domain

for example, +3 could be interpreted by looking at a case with four containers in it


and seeing that with one container removed, the case would satisfy the condition
specified by +3. Here, the quantitative properties of objects are in the physical
situation, where the goal of having a case either with a certain number of containers
or a certain number of empty places caused attention to be given to numbers of
containers or empty places in a case to find opportunities to satisfy the goal with only
a few physical exchanges.
The young salesperson in Carraher et al.'s (1985) study who calculated the price
of 10 coconuts also illustrates flexibility in finding compositions of quantities,
perhaps recognizing that each unit of 3 coconuts had a price that was convenient for
mental calculation because it was near 100 crusados. This example can be inter-
preted with a hypothesis of a mental model that simulates an activity involving
quantities. The salesperson might have simulated construction of an amount of
money corresponding to the price of 10 coconuts. I hypothesize that initially, a
mental model was formed with an object that represents the quantity 10 coconuts.
The price of 3 coconuts might have been known or might have been calculated
covertly. In any case, an object representing 3 coconuts, probably understood as part
of the 10 coconuts and associated with the quantity 105 crusados, was apparently
put into a mental model. This was replicated, and the two objects together,
representing 3 coconuts and 3 more coconuts, were associated with the combined
amounts of money, 210 crusados. The 6 coconuts now accounted for were compared
with the 10 coconuts specified in the problem, with the inference that 4 more
coconuts had to be accounted for. Then one more 3-coconut object was added to the
model, with its 105 crusados, giving a total of 315 crusados, and the problem was
completed by adding the remaining coconut of the 10 and a quantity of money, 35
crusados, to the price.
An individual can have a strong sense of quantities without being aware of
significant relations between the numbers that are involved. An example was
included in observations made by de la Rocha (1986; also discussed by Lave,
Murtaugh, & de la Rocha, 1984) in a study of recent recruits to the Weight Watchers
dieting program. One dieter was asked what would happen if he decided to use three-
fourths of his day's allotment of cottage cheese, which was two-thirds of a cup. His
solution was to measure two-thirds of a cup of cheese, drop it on the counter, pat it
into a circle, divide it into quarters with a horizontal and a vertical line, and remove
one of the quarters. This observation has been cited often by cognitive scientists
(e.g., Brown, Collins, & Duguid, 1989) because it illustrates very clearly a process
of reasoning that interacts with, and depends on, the features of the situation in
which it happens.
It is interesting, too, to reflect on aspects of number sense that the dieter example
does not show. If the relation between quantities and rational numbers was evident
to the dieter, he might have recognized that /4 of Y is the same as Y of %4, and that
2 of the 3 fourths are ?, or V (Magdalene Lampert and Judith Sowder have noted
this possibility in conversations.) Another possibility would be to recognize that %
is the same as 4%, and that 3 of the 4 sixths are -, or 2. According to the hypothesis

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
James G. Greeno 195

of mental-models, these
constructed a model wit
ties of those representat
tivity of multiplication
fraction with the numer

Computational estimat
objects in the domain a
example involves findin
could have objects repr
object. Recognition that
by the object for 43 is n
into 6 and 7 by division
In the problem of findi
and 7/8 can be represen
are approximately 1. Th
and (7, 8) both involve s
fraction are nearly the
in the neighborhood of
appropriate resolution w
might still be less than
with objects that repre
student might then reali
'2/13 + 7A and 2 is great
Many adults believe tha
without the need to con
properties. My own intu
explanation from the i
suggested by a hypoth
findings that animals c
proposed a hypothesis o
counters in the sense t
number of events. Assu
being
responsible for an
seems reasonable to sup
and states of such magn
acquisition of these asso
the presence of symbols
magnitude-recording un

Quantitative judgment
zance and reasoning at
lems like those Moore (
tudes of several quantiti
in the first section of th

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
196 Number Sense as Situated Knowing in a Conceptual Domain

average density of deciduous trees per unit of area, and the average number of leaves
per tree. Someone might know one or more of these numbers, but most people infer
them from magnitudes of other quantities that they know or can approximate. A
mental model that could support these inferences would have a representation of
North America, focusing on the approximate shape of the continent. Some numeri-
cal magnitudes may be known, at least approximately, such as the distance between
the east and west coasts of the United States, and these would be used, with spatial
information in the model, to arrive at values of other numerical values. Most of
Moore's protocols were consistent with this hypothesis, although there were some
exceptions in which problem solvers used formulas that they knew.
In the negative example from NAEP that Schoenfeld (1988) noted, successful
reasoning could be supported by a mental model. One such model would have a
representation of a large group of soldiers, with 1128 soldiers as a quantitative
property of the group. The model would also have a representation of a group of
buses, with a capacity of 38 soldiers as a quantitative property of each bus. The
distribution of soldiers into the buses could be represented in the model, correspond-
ing to the division of 1128 by 38. Representations of quantities in the model would
involve different constraints on the decomposition of 1128 into factors from those
involved if only numbers were represented. Specifically, representation of the
quantities would include a constraint that the number of buses has to be an integer,
and it has to be large enough for all the subsets of soldiers, including the subset that
remains after sets of 38 have been distributed to buses. With this constraint, the
quotient of 31 obtained when 1128 is divided by 38 is interpreted as a number of
buses that can hold all except the number that is obtained as the remainder, and
another bus is needed to hold those soldiers.

Two Further Issues

Knowing notations. Knowing in the domain of numbers and quantities also


includes fluency in the notations of arithmetic, and this is also informed by the
environmental metaphor. Someone who knows an environment well can see
through a map of that environment, understanding relations between places in the
environment that are represented by the symbols on the map. Mathematical symbols
and the concepts and relations that they refer to operate in a similar manner. We often
think of notations of place-value positions, decimal points, fraction bars, and ratio
colons as impediments to students' understanding, and they often are. At the same
time, as Hiebert (1989) pointed out, the physical properties of symbols can be
resources for reasoning-in other words, the physical symbols provide affordances
for activities of making inferences. The positions of digits in an integer or the
location of a decimal point, a fraction bar, or a ratio colon conveys important
information about the relations among numbers or quantities that are denoted by the
numerical symbols. They are also crucial cues in the performance of procedures on
the symbolic representations that include them. Graphs, too, are a problem for
students in their learning of mathematics. But graphs also present affordances for
reasoning, providing compact representations that present properties of functions
in visible form.

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
James G. Greeno 197

Uses of symbolic nota


example, noted by Carr
tation of multidigit num
different place position
example is involved in co
5 + 3 by implicitly callin
the right thumb, "six, s
with a finger simultan
cardinality of the set of
1987). For example, the n
five fingers on the left
counting operation will
ings of a single term se
system in their numeric
may play an important
differences in meaning
allowing efficient mani
pretation in different p
Processes of learning t
probably require coordin
cal quantities or mental
mental models, represe
sponding to the magnit
learning to construct an
tations of quantitative p

The social nature of re


quantities as a conceptu
of mathematical reason
groups of individuals to
can express their under
representations, thereb
environment and find n
the locations and routes
aroundrepresentations i
symbols and enrichmen
understood.

The metaphor is quite suggestive regarding the role of a teacher. As learning is


analogous to acquiring abilities for finding one's way around in an environment,
teaching is analogous to the help that a resident of the environment can give to
newcomers. Someone who already lives in the environment is an important resource
for a newcomer, helping by indicating what other resources the environment has,
where they can be found, what some of the easy routes are, and where interesting
sites are that are worth visiting. Effectiveness in providing guidance to others is not
equivalent to knowing the environment oneself; one can be fully effective in finding

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
198 Number Sense as Situated Knowing in a Conceptual Domain

and using the resources of an environment but be of little help to someone else. An
effective guide for learners needs to be sensitive to the information they already
have, to connect new information to it, to provide tasks and instructions that can be
engaged in productively by beginners, to be aware of potential errors that can result
from newcomers' partial knowledge, and to help beginners use errors as occasions
for learning. In addition to guiding students productively in their learning of the
environment, teachers also need the ability to cause students to want to participate
in the tour.

Teachers and students work together on the construction of mathematical know-


ing, but of course they do not construct mathematics di nu6vo. The mathematical
environment that students can learn to live in corresponds to the intellectual
practices of communities that have constructed the concepts and methods of
mathematics, science, engineering, and business. The main features of the math-
ematical environment have been constructed by professional mathematicians, and
students and teachers work on learning to find the resources, routes, and paths that
are there and on how to use the resources in reasoning and understanding. This does
not prevent occasional new constructions, in which a new path is discovered or a
new doorway or window is built, and these occasions are particularly valuable for
students and their teachers when they occur.
The social setting of learning mathematical concepts includes significant use of
mathematical notation. The discourse around those notations can be crucial in

determining whether students acquire capabilities of constructing and reasonin


with mental models. A too-frequent outcome of school learning is the acquisit
of procedures for manipulating notations without adequate understanding of th
meanings. In the hypothesis of mental models, understanding occurs throug
constructing and reasoning with mental models based on the mathematical symbol
A central issue for research, on this view, will be to investigate processes of ment
modeling and the conditions in which students learn to construct mental mod
rather than simply learn to manipulate notations.
The view presented here emphasizes the need for students to learn to constru
models. This suggests a different use of manipulative materials than the one that
often considered. The role of manipulative materials is often to provide studen
with models in which they can see the properties and relations that are required t
understand a concept. An alternative view is that materials should be provided
that students can construct models, thereby learning the constructive processes th
they need in order to construct mental models that can constitute understanding.
It is likely that for most students, appropriate social activities are required f
learning to construct mental models as interpretations of mathematical symbo
The meanings that notations have for most students are those that are construc
in their discourse activities. Perhaps by engaging in conversations about the
conceptual entities of mathematics, the processes of reaching mutual understandin
include constructing models that each participant uses to provide referents for ter
in the conversation. That process could be an effective means of learning to
construct mental models for many students. In current educational practice, s
dents do not participate very much in conversations about conceptual entitie

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
James G. Greeno 199

Apparently, the proces


sufficient for some stud
models with mental ob
reciprocal processes of
educational means by w
understanding and capa

RELATED RESEARCH AND THEORY

The view of number sense as knowing for living in a conceptual en


congruent with several trends in research on cognition and learning

Situated Cognition
Most generally, the environmental metaphor fits with an effort that
in cognitive research to develop an alternative to the framework of
processing. The analytical resources that have been available in the
information processing (Newell & Simon, 1972) and the behavioristic
preceded it (Gagne, 1965) enable the dissection of knowledge a
processes into components. An information-processing analysis ass
person constructs a representation of the situation and her or his go
by manipulating the symbols in the representation. The person
includes information structures corresponding to concepts and pr
Knowledge also includes procedures that make inferences and set n
constructing additions to the representation of the situation. Analysi
cognitive capability, in this framework, involves hypothesizing a set
for representing situations; a set of knowledge structures-most freq
form of schemata-that provide organization for the information and
processes; and a set of processes that make inferences of various kin
setting new goals that become appropriate as the reasoning process
information-processing framework has been used productively in
several instructional tasks used in mathematics teaching, including
computation (Brown & Burton, 1978), elementary word problems (Ri
1988), and geometry proof exercises (Greeno, 1978; Anderson,
interesting instructional methods and systems have been developed u
(Anderson, Boyle, & Yost, 1985; Carpenter, Fennema, Peterson, Ch
1989; Thompson, 1989).
The alternative view assumes that cognition is situated in contex
perception and reasoning as relations between cognitive agents and
physical situations they are in (see, e.g., Brown, Collins, & Duguid, 1
in press; Greeno, 1989a; 1989b; Lave, 1988; Suchman, 1987; Winogr
1986). Following ideas of Heidegger (1926/1962), in this view cogni
typically depend on representations, although the ability to const
representations plays an important role in some reasoning. The pr
symbolic representations is in communication, when people use lang
symbolic systems to coordinate their activity and reach shared und
events and ideas. Symbolic representations are often constructed by

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
200 Number Sense as Situated Knowing in a Conceptual Domain

provide physical objects that are useful in reasoning, such as road maps, musical
scores, diagrams of physical or mathematical systems, shopping lists, and notes to
use in giving a lecture. Cognitive representations, in the alternative view, are mental
versions of physical representations, such as mental images or silent speech, and
although they can play an important role in reasoning they are by no means
ubiquitous.
Several lines of evidence and theoretical work have encouraged the development
of the alternative view of cognition as a relation between agents and situations.
Some of the evidence regarding cognition is in studies of everyday cognition,
including Carraher et al. (1985), de la Rocha (1986), and Scribner (1984), men-
tioned earlier, and also including work by Lave (1988), Saxe (1990), and Suchman
(1987).
Some reasoning can be understood best by assuming that an individual is
connected cognitively with objects in an environment that are not immediately
present. Then the connection with those objects has to be mediated, but the
mediation is through locally available information, rather than through a mental
representation. An example is navigation by Micronesian seamen (Hutchins, 1983),
who determine their courses without maps, on the basis of the information in the
direction and speed of waves as the boat moves through the water. The navigator
keeps track of his position in reference to islands that serve as landmarks but are
generally not visible; indeed, on some journeys a landmark island is used that does
not exist. The navigator knows where the islands are relative to his boat and keeps
an updated understanding of his position as the islands move past the boat during
the journey.
The view of number sense presented in this paper is an attempt to characterize
conceptual knowledge in the framework of situated cognition. The basic form of
situated cognition is an interaction of an agent within a situation, with the agent
participating along with objects and other people to co-constitute activity. The
agent's connection with the situation includes direct local interaction with objects
and other people in the immediate vicinity as well as knowing where he or she is in
relation to more remote features of the environment. According to the hypothesis of
this paper, number sense is an example of a kind of conceptual reasoning that is
analogous to the reasoning we do in situations. We construct mental models that
provide us with situations in which we can interact with mental objects that
represent objects, properties, and relations and that behave in ways that simulate the
objects, properties, and relations that our models represent. We also know where
these models are, and where we are as we reason by interacting in the models, in the
larger conceptual environments in which the models are constructed. The concepts
and principles that a person understands, in this sense, are embedded in the kinds
of objects that he or she includes in mental models and in the ways in which those
objects behave, including how they combine and separate to form other objects and
how they are interrelated by proximity and path connections in the conceptual
domain. I hypothesize that abilities to construct mental models arise fundamentally
from our activities in conversations in which mental models are constructed in order

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
James G. Greeno 201

to provide referents for


are represented mental
capabilities in conversat
Learning also occurs as
problems, but reliance on
material misses a social l
of students.

Perceiving and cognizing affordances. The view that I have presented of concep-
tual knowing and reasoning with mental models is related to some general issues in
the psychology of perception and action that are important for a theory of situated
cognition. The ideas about number sense in this paper are consistent with some
general hypotheses that differ from hypotheses that are often made currently in
cognitive science.
The idea of conceptualizing activities as interactions in which agents and
situations are coparticipants has been proposed by several psychologists, including
Dewey and Bentley (1949) and J. J. Gibson (1986), contributors to a volume edited
by Pervin and Lewis (1978), and recent contributors to ecological psychology, such
as the contributors to McCabe and Balzano's (1986) volume. Following Gibson, we
use the term affordance to designate the contributions that things in the situation
make to the interactions. Standard examples of positive affordances include hori-
zontal surfaces that afford support, openings in walls such as doorways that afford
walking from one room to another, and stairsteps that afford climbing. Affordances
are relational; their support of activities is relative to properties of the person or
animal that engages in the activity. For example, a solid floor affords support for a
person, whereas the surface of a pond does not, although it does for some insects.
Whether an opening in a wall affords walking through it depends on its width in
relation to the size of the person, and whether a set of stairsteps affords climbing
depends on the height of the stairs in relation to the length of the person's legs.
Several experiments have investigated perception of affordances. E. J. Gibson
and Walk (1960; Walk & E. J. Gibson, 1961 ) showed that babies and other terrestrial
animals distinguished between visual patterns of a"visual cliff" and a"visual step,"
produced by textured paper under a glass floor either far below the glass or a short
distance. Babies and other terrestrial animals were placed on a section where the
visual information specified support, and almost all of them chose to move across
the visual boundary at which the apparent dropoff was shallower. Perception of
affordances for climbing stairs and for walking through doorways has also been
studied experimentally, with the conclusion that the information that specifies the
affordance is the ratio of the stair height or the width of the aperture to the eye level
of the person (Warren, 1984; Warren & Whang, 1987).
The process of perceiving affordances involves an important theoretical distinc-
tion between direct and mediated perception. J. J. Gibson (1966, 1986) argued that
information that specifies affordances for orientation and locomotion in the spatial
environment is perceived directly, rather than being constructed out of perceptions
of more elementary cues. In particular, Gibson opposed the idea that information

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
202 Number Sense as Situated Knowing in a Conceptual Domain

about the three-dimensional structure of the environment has to be constructed by


an inference process based on cues such as relative sizes of images in the two-
dimensional projection on the retina.
An elegant analysis of the information about locations of objects in space has been
given by Lee (1980). As aperson or animal moves through an environment, the point
on the retina receiving light from any given point on an object in space moves away
from the center of the retinal field in a direction and at a rate that depends on the
location of the point on the object in the environment and the speed and direction
of movement. Liebowitz and Post (1982) reviewed evidence that the human visual
system processes information in two modes, which they called thefoveal mode and
the ambient mode. The foveal mode provides information for object recognition and
identification; the ambient mode provides information for spatial orientation and
locomotion, posture, and the coordination of motor activity with the various forms
of sensory information. Neisser (1989) noted that properties of information about
spatial orientation and locomotion and information for object recognition are
different in fundamental ways; for example, the information for spatial orientation
is highly redundant, has an invariant mathematical structure, and is obtained most
effectively by a person or animal moving in an environment, whereas information
for object recognition has several dimensions that can vary independently, accumu-
lates over time, and is obtained most effectively when the person and the object are
stationary. Neisser concluded that recognition and direct perception should be
considered as distinct systems of visual perception. In recognition, features of an
object or event are detected and compared with information stored in memory in the
form of patterns of features. In direct perception, it is more accurate to consider the
person and the situation as a dynamic interactive system in which the person
responds on the basis of information without necessarily having a symbolic
representation of that information.
According to this way of thinking, affordances do not have to be recognized in
order to be perceived. For example, the information that specifies an affordance of
walking through a doorway, the ratio of the width of the doorway to one's eye level,
can be detected without recognizing a pattern of visual features such as the
intersecting vertical and horizontal lines that would correspond to a conceptual
representation of what one calls a "doorway." Similarly, I assume that if there is a
closed door with a doorknob, the affordance of turning that the doorknob presents
can be perceived directly, as can the further affordance of pulling or the affordance
of pushing when the handle is turned and the further affordance of walking through
the open space that will result when the door has been opened. Settings differ in their
presentations of affordances. Norman (1988) collected examples in which the
information in the situation misleads people, especially newcomers. One example
involves round door handles that look as though they require turning to open the
door, but in fact the door can be opened simply by pulling or pushing.
Perception of affordances is influenced by attention and by learning. The activity
that a person is engaged in influences whether a given affordance will be perceived.
For example, a stairway is designed to afford climbing, but it also affords sitting. The

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
James G. Greeno 203

affordance of a stairway f
person is walking or runn
may be perceived, howeve
a conversation or if someo
affordance for an activity
in that activity sufficiently
negative affordance of a vi
but as a more typical exam
hinged doors, it is unlikely
a door by turning a doorkn
that we perceive only aft
water faucets for getting
bottles, and mailboxes for
depends on the person's or
in the situation in the way
the situation, however, d
features of actions that on
knows how to perform the
category u of object that is
I can report an experienc
tion of affordances. A you
her geometry course to pr
sides of a parallelogram for
a diagonal of the parallelog
midpoint is par connector
initial efforts to get that
triangle formed by the d
one-half of the figure w
recognized that the line c
remembered the theorem t
perception of a feature of
person knew the pattern o
that requires a particular
An important kind of aff
together to make other th
kitchen, the presence of i

2 Neisser (1989) sided with opponent


distinguish here between direct perce
sensory information that is present i
specified in properties thatthe person
are some affordances that can be perc
affordance of removing a cork from a
that can be cognized directly in a wa
imply a total lack of symbolic represe
as being a corkscrew or a mailbox to
of the object that enable its use woul

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
204 Number Sense as Situated Knowing in a Conceptual Domain

things to eat. In a woodworking shop, the presence of materials, fasteners, and tools
affords activities of making objects such as furniture. Compositional affordances
are particularly important in the phenomena of number sense, as Trafton (1989)
observed and as I have discussed in the previous section of this paper.
Compositional affordances are generally important in constructing mental mod-
els. This is illustrated in the kind of mental model called a thought experiment. As
an example, consider Galileo's (1632/1967) discussion of the path of a ball of wax
dropping slowly in a deep vase filled with water. The vase is on the deck of a sailboat,
and to someone on the boat the path appears to be vertical. The sailboat is moving
in a large circle, and relative to the earth, the path of the ball is a compound of the
vertical motion with the motion of the boat. The mental model that supports this
reasoning includes representatives of some objects (a boat, a vase, a ball of wax),
and simulations of motions of the objects. The simulated motion of the ball through
the vase and the simulated motion of the boat can be combined, and the simulation
has the resulting property of the ball moving along a helical path.
Situations also present affordances for constructing representations of their
features that can be used in communication and reasoning. Symbolic representa-
tions are used in conversations to designate properties of objects in the situation, as
when a person remarks on an interesting or beautiful feature of a landscape or asks
another person to pass a dish during a meal. An affordance for an action in a situation
can also be an affordance for communication using a denotational representation.
In this way, affordances are informationally productive. If someone detects an
affordance and represents properties of the situation that would result from a
change, that representation provides information about the situation that might not
be available otherwise. Examples include directing individuals in a group to sit in
specified places around a table (e.g., "Tom can sit here and Sue can sit beside him
here") or agreeing on an allocation of tasks (e.g., in sawing a board, "I'll hold this
end while you do the sawing"). The spoken, written, or drawn representations
become part of the situation and present affordances for further actions and
reasoning, such as working out a complicated seating arrangement by drawing a
chart or drawing a line across a board to aid in sawing a straight edge.
Mental models provide affordances for reasoning. The mental objects that a
person includes in a model provide affordances for mental activities. These
affordances and the mental operations that they support are informationally produc-
tive. Constraints that are implicit in the behavior of objects in the model provide a
basis for making inferences. Symbolic representations of features of objects in the
model can be constructed to designate their properties and properties of the mental
operations that are performed to simulate operations that could be performed on
physical objects, and these representations communicate information about
inferences.

Constraints that apply to mental models need not be exactly those of the physical
world that we experience, but they should involve some principles for simulations
to occur with regularities that would make them useful. Some principles should be
easier to incorporate into mental models than others-perhaps principles that

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
James G. Greeno 205

correspond to phenom
therefore easier to use
principles that are em
consistent with physica
to those by principles
Conceptual entities in
reasoning in several w
can be recognized and
such as statements or
changed by operation
among properties and
conceptual entities in a
can be based on percept
recognizing a property

Conceptual Competenc

The hypothesis that w


within mental models
activities that people
Learning this kind of
tual activity that peop
capabilities for listenin
cepts.
In many domains, children's understanding grows naturally as a result of their
ordinary interactions with adults and other children along with their natural
curiosity and inclinations to learn to participate in social interactions. Young
children become skilled in speaking and understanding their language and in using
language in ways that are practiced and valued in their social and cultural environ-
ments. For example, Heath (1983) described different linguistic and communica-
tive abilities that children acquired as a result of differences in the linguistic
practices of the two communities in which they were raised. In one of the
communities children learned to tell stories that have morals in accord with
community values, usually about some event that illustrates negative consequences
of a wrongful act, with strict constraints of accuracy in statements of fact and
chronology. In the other community, children learned to participate in elaborate
expressive word play, including playsongs and ritual insults presented creatively,
and stories with real events as the bases of the plots, but with details elaborated and
exaggerated to emphasize truths about life rather than to report facts accurately.
Significant aspects of conceptual understanding also develop naturally in many
conceptual domains. Properties of knowing, understanding, and reasoning in
several domains have been investigated by developmental psychologists in studies
of conceptual growth. Knowing in a domain includes distinguishing the entities and
phenomena in the domain and knowing the principles of identity, invariance,
composition, transformation, and causality that can be used to explain phenomena

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
206 Number Sense as Situated Knowing in a Conceptual Domain

and events. It includes knowing the conceptual and theoretical entities of the
domain, the principles of invariance, composition, transformation, and causality
that apply to the theoretical entities. It also includes knowing how to use the terms
to designate properties, state principles, refer to theoretical entities, and construct
explanations of events in the domain. All of these conceptual and linguistic
capabilities develop through activities in which individuals communicate and
negotiate the meanings of terms to coordinate their actions and to achieve mutual
understanding of events.
As an example, in the domain of biological phenomena, a principle of invariance
was studied by Keil (1989). Keil showed children of different ages pictures of
different kinds of animals, such as a raccoon and a skunk, and different kinds of
artifacts, such as a bird feeder and a coffee pot. The interviewer showed the two
animal pictures and said that some scientists had changed the first animal's color,
shape of tail, and other features to those shown in the picture of a skunk. The
interviewer then asked whether it should be called a raccoon or a skunk. Similar
questions were asked about a bird feeder and a coffee pot. Young children replied
that the second animal should be called a skunk and the second artifact should be
called a coffee pot, because their features had been changed. Older children said that
if the animal was a raccoon, it should still be called a raccoon after its features were
changed, although they said that the artifact had been changed into a coffee pot. The
finding shows that older children come to understand that principles of identity for
natural kinds are different from those of artifacts.
Another example involves principles of transformation, including constraints.
Carey (1985) reported that young children answer the question "Why do people
eat?" in terms of psychological concepts of desires and feelings-people eat
because they feel hungry or they want food. Older children understand that eating
is necessary for continuing to live and for growth. Hatano and Inagaki (1987) found
that if 6-year-old children are asked whether a baby rabbit can be kept small by not
feeding it, they understand that this will cause the rabbit to die. Apparently the
younger children have some understanding of the functional principles of nutrition
and growth but do not spontaneously use these principles as explanations in some
situations where older children and adults use those biological principles.
A third example illustrates understanding of theoretical entities and processes.
Wellman and his associates (e.g., Johnson & Wellman, 1980; Wellman & Estes,
1986) have shown that children as young as three years understand the difference
between mental states such as thinking about a dog and the physical objects that the
mental states are about and that they use concepts such as a person's knowledge in
explaining behavioral events. Thoughts and states of knowledge are not observable,
and children's use of these to explain the behavior of another person illustrates the
role of unobservable conceptual entities and processes in a domain that they
understand well early in their lives.
In the domain of numbers, Gelman and Gallistel (1978) showed that children have
significant implicit understanding of counting principles that provide a beginning
of conceptual knowing in the domain of numbers and quantities. Resnick and

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
James G. Greeno 207

Greeno (1990) are wor


intuitive understanding
and intuitive understa
combination.

Case (1989) and Case and Sowder (in press) have used a theory developed by
Case and Griffin (in press) to discuss developmental aspects of computational
estimation. According to Case and Griffin's general theory, children develop from
being capable of conceptualizing a single quantitative dimension at about age 6, to
being capable of representing two dimensions and then two dimensions with
compensation at about age 8, to being able to coordinate two activities that they
could execute previously only in isolation between about ages 10 and 12. Case's
analyses and discussions indicate an important aspect of the growth of children's
capability for interacting with the conceptual domain of mathematics. As children
grow older and are more experienced with numbers, they will become able to
interact with more complex features of the domain, and interact in more complex
ways. The multilevel, multiply connected character of knowing in the domain is less
available to young children than it is as they grow older.
Another implication of the literature on conceptual competence and conceptual
growth has to do with the assumptions that are made in instruction about students'
conceptual capabilities in advance of receiving instruction and about the nature of
their learning. A common assumption that children will not understand mathemati-
cal concepts until they are taught them is too simple. They have significant implicit
understanding of many concepts and principles, and can reason on the basis of those
concepts when there are appropriate supporting features in the environment.
Children also develop more articulate and more complex understanding of concepts
through a process that is similar to processes of theory change in important ways.
Processes of instruction can be thought of as facilitating natural processes of
conceptual growth, rather than causing acquisition of only those contents that are
provided directly.

Growth of Cognitive Expertise


Another line of evidence is presented in a recent study of the nature and
acquisition of expertise in cognitive domains by Dreyfus and Dreyfus (1986).
According to an influential model in psychology (Fitts, 1964) skill acquisition has
three main stages: an initial cognitive stage in which general properties of actions
are learned by being given verbal descriptions and rules and observing performance
by others, an associative stage in which the person learns to perform actions
correctly, and then a stage of automatization in which the actions that have been
learned come to be performed rapidly without conscious control. Dreyfus and
Dreyfus's discussion is consistent with part of the standard psychological analysis,
but it also provides a new insight into the process. Dreyfus and Dreyfus agreed that
initial stages of learning a skill are guided by verbal descriptions and rules and that
translating cognitive representations into actions can get one to an intermediate
stage of skill. They disagreed, however, about the way that performance improves
toward expertise.

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
208 Number Sense as Situated Knowing in a Conceptual Domain

Rather than assuming that expertise consists of automatizing previously learned


procedures, Dreyfus and Dreyfus hypothesized that expertise includes many
capabilities that were never dependent on rules or descriptions. Rather than
assuming that automatic skill is mainly a converted form of skill that once was
controlled deliberately, they hypothesized that most of what matters in expert
performance is learned without being deliberately controlled. These nondeliberate
aspects of skill involve recognition of holistic patterns and interactive engagement
with situations including flexible adjustment to nuances and configurations that the
person does not consider consciously, and probably never considered consciously.
One example that Dreyfus and Dreyfus considered is the skill of driving a car. The
earliest stages of learning include following directions for the sequence of things to
do to start the car, rules for signalling before making a turn, and so on. A beginning
driver practices the basic activities of driving until he or she performs them without
making major violations of rules such as turning from the wrong lane or driving into
the crosswalk when stopping at an intersection. When the driver is proficient
enough, he or she can obtain a license to drive. If the person drives for several years,
however, the skill he or she acquires includes abilities to adjust the speed of the car
to allow another car to enter a lane, to identify a space between cars in another lane
he or she wants to enter, to turn corners at different speeds and on different road
surfaces so that the ride is smooth, and many other capabilities that were not taught
deliberately (although there may have been remarks made about them, in the manner
of coaching).
Dreyfus and Dreyfus studied expertise in piloting airplanes, playing chess,
driving cars, communicating in a second language, and nursing. They characterized
the features of expertise in these domains as global, holistic, and not derived from
rule-based learning. As an example, they reported that one of them had played chess
fairly seriously but had not progressed beyond a medium level of skill. In light of
the analyses that they presented in Mind Over Machine, he came to believe that his
progress was limited by treating the game too analytically, trying too much to
identify rules and principles of successful play and blocking himself from acquiring
the holistic and configural aspects of skill that are not amenable to descriptions in
abstract terms.

An important distinction regarding expertise was made by Hatano (1988)


between routine and adaptive expertise. Hatano's example was expertise in comput-
ing with the abacus, a skill that is highly developed by children who join clubs and
become highly skilled. The expertise that these children achieve goes beyond rule-
based knowledge, consistent with Dreyfus and Dreyfus's (1986) characterization of
expert knowledge. The abacus experts had mental models of the abacus that they
manipulated in flexible, problem-sensitive ways. Being situated in a conceptual
model of the abacus, however, did not provide experts with concepts of numbers and
quantities that would enable them to generate ways of understanding and reasoning
in a broad range of situations.
Dreyfus and Dreyfus's (1986) view of expertise reinforces the view that number
sense is unlikely to develop through the learning of rules for mental computation,

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
James G. Greeno 209

estimation, and quantitat


students engage in acti
salient factors, and wher
among numbers and qua

Conversation and Communication

Another set of results and analyses have been provided in studies of conversa-
tional communication (e.g., Clark & Shaeffer, 1989; Clark & Wilkes-Gibbs, 1986;
Schegloff, 1981). Conversation has typically been considered as a series of turns,
in which a speaker presents a message and the listener(s) either do or do not
understand the message's meaning. The more recent analyses take another view,
that meanings are achieved jointly by the speaker and listener(s) through a process
of collaboration and negotiation. Clark and Shaeffer drew an important distinction
between presenting information and contributing information to the conversation.
Presenting is done by an individual. Contributing is a collaborative act that is
accomplished only when the speaker and listener(s) agree that something meaning-
ful has been added to the collection of information that is shared in their common
ground. In this view, meanings of words and phrases are not fixed, to be known or
not by participants. Rather, meanings of terms are part of what emerges in the
conversational context as the participants develop uses of terms that support their
communicative activities. Of course, conversations do not start at ground zero with
respect to word meanings, because members of a linguistic community share most
of the terminological agreement that they need to communicate successfully in most
situations. But a process of semantic negotiation occurs to some extent in all
situations, and to a significant extent in many situations, especially when some
participants in the conversation have less experience and familiarity with the topic
than others.

Much of our ability to understand each other in conversations depends on our


being in the same situation and sharing access to the objects and properties that we
are talking about. Stucky (1989) has noted that an important aspect of understanding
in conversations involves the ability of different persons to attend to the same
objects and properties in the situation. For example, if someone at a dinner table asks
another person, "Pass the asparagus, please," they can both tell easily what dish is
meant if there is just one dish of asparagus there, they both can see it, and they both
are able to recognize asparagus. In a more complicated example, an air traffic
controller, conversing with a pilot, may refer to "the plane on your left." The
instruction that the controller gives to the pilot probably depends on their both
knowing which plane the controller means to refer to. This probably could not
happen if the pilot and the controller did not have access to information about the
same situation-the pilot directly, by having a plane to his or her left, and the
controller indirectly, by having a representation of the planes flying in the area.
However, both the controller and the pilot understand the phrase as referring to a
physical airplane, not a blip on the controller's radar screen.
Much of our communication involves information about situations that we are not
in. Faucconier (1985) and Johnson-Laird (1983) have developed theories of

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
210 Number Sense as Situated Knowing in a Conceptual Domain

language understanding based on the idea that we construct mental situations,


including conceptual objects corresponding to the individuals, objects, and events
that the language is about. If someone tells a story at home to his or her spouse about
an event that occurred at work, the speaker and listener construct a mental situation
containing the workplace, the people in the story, and the objects and events that are
involved. When one reads fictional literature or a newspaper report of an event
somewhere in the world, according to Faucconier's and Johnson-Laird's hypoth-
eses, the reader constructs a mental situation that constitutes his or her understand-
ing of the story.
Much of our mathematical activity takes place in conversations and reading of
text. The idea of situated cognition and language suggests that understanding the
language of mathematics depends on developing the capability of constructing
mathematical situations that include the conceptual entities that the language is
about. Much remains to be understood about processes of developing the capability
of constructing mathematical situations in the understanding of mathematical
language. It seems clear, however, that the common practice of introducing a term
by means of a formal definition and a few examples probably does not suffice to
make that term effective in most students'processes of constructive understanding.
The idea that meanings are negotiated in the contexts of conversations, rather than
being carried autonomously by terms in the language, has significant implications
for the character of conversations between teachers and students. In mathematical
conversations, the intended referents of many terms are mathematical objects,
which students must learn to construct, rather than physical objects, which are
present in the situation. To have a successful conversation, the participants construct
common ground-that is, a shared body of understanding-and conversational
processes are available to check whether the participants' understandings are in fact
shared and to repair mismatches that occur. An important problem for research and
practice is to come to understand how conversational checks and repairs can
function effectively when the referents of terms involve the abstract concepts and
complex relations among the quantities that figure in mathematical conversations.

Intellectual Practices, Cognitive Apprenticeship, and Personal Epistemologies

The idea of learning in a subject-matter domain by being guided by someone who


already is familiar with the conceptual environment has much in common with a
view of learning through cognitive apprenticeship (Brown, Collins, & Duguid,
1989; Collins, Brown & Newman, 1989; Lave & Wenger, in press; Rogoff, 1990).
They emphasized that knowing in a domain is an activity, and learning in the domain
is acquiring the capabilities of understanding and reasoning that the domain affords,
a kind of practice.
A characterization of the practice of mathematics was provided by Kitcher (1984)
in a philosophical and historical analysis of mathematical knowledge. Knowing in
the domain of mathematics, according to Kitcher's idea, includes (a) ability to
understand and generate meaningful mathematical questions; (b) ability to appre-
ciate and use methods of mathematical reasoning that are accepted as supporting

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
James G. Greeno 211

conclusions; (c) understa


and structures of math
mathematical language;
conclusions that are ac
outcome of general kno
contribute to the underst
Recent research and di
(1986) and by Dweck an
the beliefs that individua
as knowers and learners,
learning. The idea of l
environment emphasizes
logical view in which th
own, as well as those of o
information. We learn ab
travelogues and other ex
teach us how to live there
in learning activ through
The view of knowing a
to the growth of number
with number sense go be
pation in activities. As R
number sense when they
to the surface of intellec
islikely to foster childre
numbers, as Resnick (198
encouraging positive ep
participation in the activ

Relations With Other Theories

Information-processing. The hypothesis that I propose here competes directly


with the information-processing view of conceptual understanding and reasoning,
in which definitions of concepts are represented in schematic networks of proposi-
tional structures and reasoning occurs by constructing and manipulating notational
expressions that are interpreted as referring to objects and properties in the situation.
In place of the network of schemata I propose a conceptual environment with spatial
properties, and in place of propositional expressions I propose construction of
mental models and reasoning as a process of interacting within the models that are
constructed.
The schema/proposition and the environment/model views have much in com-
mon. Both hypotheses focus on relations among concepts and information that is
available for reasoning. The views differ primarily in the forms of information that
are learned and the kinds of mental activity that constitute reasoning.
In the schema/proposition hypothesis, learning is analogous to writing text into
long-term memory in a format that provides an organized set of information

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
212 Number Sense as Situated Knowing in a Conceptual Domain

structures that can be retrieved efficiently, acquiring abilities to recognize situations


that match stored information, interpret inferred propositions, and perform actions
that correspond to propositional instructions. Reasoning includes writing text in
working memory based on stimulus inputs and on information that is retrieved from
long-term memory, applying rules that generate additional text, and interpreting the
generated text as instructions for action.
In the environment/model hypothesis, learning is the acquisition of behavioral
abilities, analogous to knowing how to assemble and disassemble things, knowing
how to find the ingredients needed for these assembling and disassembling
activities in an organized environment, and knowing how to interact with the things
that are constructed to see what happens when various operations are applied.
Neither hypothesis excludes the processes that the other hypothesis considers
basic. In the schema/proposition view, alternative representational formats such as
imagery are often assumed as a kind of supplement to the propositional mode. In the
environment/model view, written representations, drawings, and other physical
representations are parts of the situations within which people interact. People learn
to interact with physical symbols as they do with other physical objects, with the
addition that they learn to interpret symbols as denoting other objects and properties.
It is possible, as well, to learn to construct mental models in which the objects are
representations of physical symbols and their interactions within these mental
models simulate operations on physical notations. These cognitive denotational
representations are a kind of supplement to the basic mode of simulative reasoning.
These two views differ on general theoretical grounds as well as in the focus that
they provide for teaching and learning concepts. In the schema/proposition view,
knowledge of concepts is separate from activities that use them. Abstract knowledge
is primary, and modeling is an auxiliary aspect of conceptual knowledge. Modeling
activity is used in teaching as a means to learning, and may be used as a test of
students' ability to apply concepts that they know. We assess students' learning
mainly by asking them to tell us about concepts.
In the environment/model view, modeling is primary and descriptions of concepts
play an auxiliary role. The main capability that we want students to acquire involves
constructing and reasoning within models. We want that activity to become
internal-they should not have to depend on having physical props in order to think
about numbers and quantities. Conceptual reasoning, however, is mainly a mental
version of concrete activity in which the symbols used to represent things,
properties, and relations behave in ways that support reasoning. Descriptions of
concepts can undoubtedly be a helpful aid to acquiring modeling capabilities that
support correct reasoning, but they do not constitute the most important aspect of
conceptual knowledge in this view.

Relations with general developmental theories. I believe that this view of con-
ceptual learning is consistent with important aspects of theoretical positions that are
influential in research in mathematics education and in developmental psychology.
The modeling view shares with Piaget's theory an emphasis on learning that is
grounded in children's cognitive activity. It shares with Vygotsky's theory an

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
James G. Greeno 213

emphasis on the social p


from the more solipsist
structures are coconstru
that people communicat
Vygotsky, in its claim t
operational activity.

REFERENCES

Anderson, J. R. (1982). Acquisition of cognitive skill. Psychological Review, 89, 396-4


Anderson, J. R., Boyle, C. F., & Yost, G. (1985). The geometry tutor. In A. Joshi (Ed.), P
the Ninth International Joint Conference on Artificial Intelligence (pp. 1-7). Los Altos,
Kaufman.

Ayers, T., Davis, G., Dubinsky, E., & Lewin, P. (1988). Computer experiences in learning composition
of functions. Journal for Research in Mathematics Education, 19, 243-259.
Behr, M. J. (1989). Reflections on the conference. In J. R. Sowder & B. P. Schappelle (Eds.). Establishing
foundationsfor research on number sense and related topics.: Report of a conference (pp. 85-88). San
Diego, CA: Center for Research in Mathematics and Science Education, San Diego State University.
Belenky, M. F., Clinchy, B. M., Goldberger, N. R., & Tarule, J. M (1986). Women's way's of knowing.
New York, NY: Basic Books.

Beth, E. W., & Piaget, J. (1966). Mathematical epistemology and psychology. Dordrecht: Reidel.
Blackwell, D., & Henkin, L. (1989). Mathematics: Report (f the Project 2061 Phase I Mathemati
Panel. Washington, DC: American Association for the Advancement of Science.
Brown, J. S., & Burton, R. B. (1978). Diagnostic models for procedural bugs in basic mathematical skil
Cognitive Science, 2, 155-192.
Brown, J. S., Collins, A., & Duguid, P. (1989). Situated cognition and the culture of learnin
Educational Researcher, 18, 32-42. (Also available as IRL Report 88-0008, Palo Alto CA: Institut
for Research on Learning)
Carey, S. (1985). Conceptual change in childhood. Cambridge MA: MIT Press-Bradford Books.
Carpenter, T. P. (1989). Number sense and other nonsense. In J. T. Sowder & B. P. Schappelle (Eds
Establishing foundations for research on number sense and related topics: Report of a conference (pp.
89-91). San Diego, CA: Center for Research in Mathematics and Science Education, San Diego State
University.
Carpenter, T. P., Fennema, E., Peterson, P. L., Chiang, C.-P., & Loef, M. (1989). Using knowledge of
children's mathematics thinking in classroom teaching: An experimental study. American Educa-
tional Research Journal, 26(4), 499-53 1.
Carraher, T., Carraher, D., & Schliemann, A. (1985). Mathematics in the streets and in schools. British
Journal of Developmental Psychology, 3, 21-29.
Case, R. (1989). Fostering the development of children's number sense: Reflections on the conference.
In J. T. Sowder & B. P. Schappelle (Eds.), Establishingfoundationsfor research on number sense and
related topics: Report of a conference (pp. 57-64). San Diego, CA: Center for Research in Math-
ematics and Science Education, San Diego State University.
Case, R., & Griffin, S. (in press). Child cognitive development: The role of central conceptual structures
in the development of scientific and social thought. In C. A. Hauert (Ed.),_A, dvances in psychology--
developmental psychology:. Cognitive, perceptuo-motor and neurological perspectives.
Case, R., & Sowder, J. T. (in press). The development of computational estimation: A neo-Piagetian
analysis. Cognition and Instruction.
Chase, W. G. (1983). Spatial representations of taxi drivers. In D. R. Rogers & J. A. Sloboda (Eds.), The
acquisition of svmbolic skills (pp. 391-411). New York: Plenum Press.
Clancey, W. J. (in press). The frame of reference problem in the design of intelligent machines. In K.
VanLehn (Ed.), Architectures for intelligence. Hillsdale, NJ: Lawrence Erlbaum Associates.
Clark, H. H., & Shaeffer, E. F. (1989). Contributing to discourse. Cognitive Science, 13, 259-294.

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
214 Number Sense as Situated Knowing in a Conceptual Domain

Clark, H. H., & Wilkes-Gibbs, D. (1986). Referring as a collaborative process. Cognition, 22, 1-40.
Clement, J. (1989). Learning via model construction and criticism. In G. Glover, R. Ronning, & C.
Reynolds (Eds.), Handbook ofcreativity. Assessment, theory, and research (pp. 341-38 1). New York,
NY: Plenum.

Clement, J., & Barowy, W. (1980). Kinesthetic representations in problem solving. Department of Physic
and Astronomy, University of Massachussets.
Cobb, P., & Merkel, G. (1989). Thinking strategies: Teaching arithmetic through problem solving. In
R. Trafton & A. P. Shulte (Eds.), New directions for elementary school mathematics (pp. 70-8
Reston, VA: National Council of Teachers of Mathematics.
Collins, A., Brown, J. S., & Newman, S. E. (1989). Cognitive apprenticeship: Teaching the craft
reading, writing, and mathematics. In L. B. Resnick (Ed.), Knowing, learning, and instruction.: Essays
in honor of Robert Glaser. Hillsdale, NJ: Lawrence Erlbaum Associates.
Confrey, J. (April, 1990). Origins, units, and rates: The construction ofa splitting structure. Paper presen
at the American Educational Research Association, Boston, MA.

de la Rocha, O. (1986). Problems of sense and problems of scale.: An ethnographic Study of arithmetic
in everyday life. Doctoral dissertation, University of California, Irvine.
Dewey, J., & Bentley, A. F. (1949). Knowing and the known. Boston, MA: Beacon Press.
diSessa, A. A. (1983). Phenomenology and the evolution of intuition. In D. Gentner & A. Stevens (Ed.),
Mental models (pp. 15-33). Hillsdale, NJ: Lawrence Erlbaum Associates.
Dreyfus, H. L., & Dreyfus, S. E. (1986), Mind over machine. New York: The Free Press.
Dweck, C., & Legett, E. L. (1988). A social-cognitive approach to motivation and personality.
Psychological Review, 95, 256-273.
Einstein, A. (1961). Relativity: The special and general theory. New York, NY: Crown.
Faucconier, G. (1985). Mental spaces. Aspects of meaning construction in natural language. Cambridge,
MA: MIT Press-Bradford Books.

Fennema, E., Carpenter, T., Keith, A., & Jenkins, M. (March, 1989). Cognitively guided instruction.
Presentation at the annual meeting of the American Educational Research Association, San Francisco,
CA.

Fitts, P. M. (1964). Perceptual-motor skill learning. In A. W. Melton (Ed.), Categories of human


learning. (pp. 243-285). New York, NY: Academic Press.
Forbus, K. D., & Gentner, D. (1983). Learning physical domains: Towards a theoretical framework.
Proceedings of the 1983 machine learning workshop (pp. 198-202). Monticello, IL.
Fuson, K. C. (1988). Children's counting and concepts of number. New York, NY: Springer-Verlag.
Gagn6, R. (1965). The conditions of learning. New York, NY: Holt, Rinehart, & Winston.
Galileo (1967). Dialogue concerning the two chief world systems. Berkeley, CA: University of Cali-
fornia Press. (Originally published in 1632).
Gallistel, C. R. (1990). The organization of learning. Cambridge, MA: MIT Press-Bradford Books.
Gelernter, H. (1963). Realization of a geometry-theorem proving machine. In E. A. Feigenbaum & J.
Feldman (Eds.) Computers and thought (pp. 134-152). New York, NY: McGraw-Hill.
Gelman, R., & Gallistel, C. R. (1978). The child's understanding of number. Cambridge, MA: Harvard
University Press.
Gentner, D., & Stevens, A. L. (Eds.) (1983). Mental models. Hillsdale, NJ: Lawrence Erlbaum Associates.
Gibson, E. J., & Walk, R. D. (1960). The visual cliff. Scientific American, 202, 64-71.
Gibson, J. J. (1966). The senses considered as perceptual systems. Boston, MA: Houghton Mifflin.
Gibson, J. J. (1986). The ecological approach to visual perception. Hillsdale, NJ: Lawrence Erlbaum
Associates.

Greeno, J. G. (1978). A study of problem solving. In R. Glaser (Ed.), Advances in instructional psy-
chology (vol. 1, pp. 13-75). Hillsdale, NJ: Lawrence Erlbaum Associates.
Greeno, J. G. (1989a). A perspective on thinking. American Psychologist. 44, 134-141. (Also available
as IRL Report 88-0010, Palo Alto, CA: Institute for Research on Learning.)
Greeno, J. G. (1989b). Situations, mental models, and generative knowledge. In D. Klahr & K. Kotovsky

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
James G. Greeno 215

(Eds.), Complex information p


Erlbaum Associates. (Also avai
Learning.)
Greeno, J. G., Riley, M. S., & Gelman, R. (1984). Conceptual competence and young children's
counting. Cognitive Psychology, 16, 44-143.
Hanson. N. R. (1961). Patterns of discovery. Cambridge, UK: Cambridge University Press.
Harel. G., & Kaput, J. J. (in press). The role of conceptual entities in building advanced mathematical
concepts and their symbols. In D. Tall (Ed.), Advanced mathematical thinking. Dordrecht, Holland:
Reidel.

Hatano, G. (1988). Social and motivational bases for mathematical understanding. In G. B. Saxe, & M.
Gearheart (Eds.), Children's mathematics (pp. 55-70). San Francisco, CA: Jossey-Bass.
Hatano, G., & Inagaki, K. (1987). Everyday and school biology: How do they interact? Quarterly
Newsletter of the Laboratory of Comparative Human Cognition, 9, 120-128.
Heath, S. B. (1983). Ways with words. Cambridge, UK: Cambridge University Press.
Hegerty. M., Just, M. A., & Morrison, I. R. (1988). Mental models of mechanical systems: Individual
differences in qualitative and quantitative reasoning. Cognitive Psychology, 20, 191-236.
Heidegger, M. (1962). Being and time. New York, NY: Harper & Row. (Original work published in
German, 1926).

Hesse. J. J, III, & Anderson, C. W. (in press). Students' conceptions of chemical change. JournalofResearch
ini Science Teaching.
Hiebert, J. (1989). Reflections after the conference on number sense. In J. R. Sowder & B. P. Schappelle
(Eds.). Estahlishing foundations Jr research( on number sense and related topics: Report of a
conference (pp. 82-84). San Diego, CA: Center for Research in Mathematics and Science Education,
San Diego State University.
Hope, J. (1989). Promoting number sense in school. Arithmetic Teacher, 36(6), pp. 12-16.
Hope, J. A., & Sherrill, J. M. (1987). Characteristics of unskilled and skilled mental calculators. Journal
for Research in Mathematics Education, 18, 98-111.
Howden. H. (1989). Teaching number sense. Arithmetic Teacher, 36(6), 6-11.
Hutchins, E. (1983). Understanding Micronesian navigation. in D. Gentner & A. L. Stevens (Eds.),
Mental models (pp. 191-225). Hillsdale, NJ: Lawrence Erlbaum Associates.
Johnson, C. N., & Wellman, H. M. (1980). Children's developing understanding of mental verbs:
Remember, know, and guess. Child Development, 51, 1095-1102.
Johnson-Laird, P. N. (1983). Mental models: Towards a cognitive science of language, inference, and
consciousness. Cambridge, MA: Harvard University Press.
Jones, W. T. (1952). A history of western philosophy. New York, NY: Harcourt, Brace.
Kaput, J. (1987). Toward a theory of symbol use in mathematics. In C. Janvier (Ed.), Problems of
representation in mathematics learning and problem solving (pp. 159-196). Hillsdale, NJ: Lawrence
Erlbaum Associates.

Keil, F. (1989). Concepts, kinds, and cognitive development.. Cambridge, MA: MIT Press-Bradford
Books.

Kitcher, P. (1984). The nature of mathematical knowledge. New York, NY: Oxford University Press.
Lampert, M. (1986). Knowing, doing, and teaching multiplication. Cognition and Instruction, 4, 305-
342.

Lampert, M. (1990). When the problem is not the question and the solution is not the answer:
Mathematical knowing and teaching. American Educational Research Journal, 27, 29-64.
Lave, J. (1988). Cognition in practice: Mind, mathematics, and culture in everyday life. Cambridge, UK:
Cambridge University Press.
Lave, J., Murtaugh, M., & de la Rocha, O. (1984). The dialectic of arithmetic in grocery shopping. In
B. Rogoff & J. Lave (Eds.), Everyday cognition: Its development in social context. (pp. 95-116).
Cambridge, MA: Harvard University Press.
Lave, J., Smith, S., & Butler, M. (1988). Problem solving as everyday practice. In R. I. Charles & E. A.
Silver (Eds.), The teaching and assessing of mathematical problem solving (pp. 61-8 1). Reston, VA:
National Council of Teachers of Mathematics.

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
216 Number Sense as Situated Knowing in a Conceptual Domain

Lave, J., & Wenger, E. (in press). Legitimate peripheral participation. New York, NY: Cambridge
University Press.
Lee, D. N. (1980). The optic flow field: The foundation of vision. Philosophical Transactions of the
Royal Society of London, 290, 169-179.
Liebowitz, H. W., & Post, R. B. (1982). The two modes of processing concept and some implications.
In J. Beck (Ed.), Organization and representation in perception. Hillsdale, NJ: Lawrence Erlbaum
Associates.

Lindsay, R. K. (1988). Images and inference. Cognition, 29, 229-250.


Markovitz, Z. (1989). Reactions to the number sense conference. In J. T. Sowder & B. P. Schappelle
(Eds.), Establishing foundations for research on number sense and related topics.: Report of a
conference (pp. 78-81 ). San Diego, CA: Center for Research in Mathematics and Science Education,
San Diego State University.
McCabe, V., & Balzano, G. J. (Eds.) (1986). Event cognition: An ecological perspective. Hillsdale, NJ:
Lawrence Erlbaum Associates.

McCloskey, M. (1983). Naive theories of motion. In D. Gentner & A. Stevens (Eds.), Mental models (p
299-324). Hillsdale, NJ: Lawrence Erlbaum Associates.
Moore, J. L. (1987). Back-of-the-envelope problems (Report No. GK-3). School of Education, Univer-
sity of California, Berkeley.
Morrison, P. (1963). Fermi questions. American Journal of Physics, 31, 626-627.
National Council of Teachers of Mathematics (1989). Curriculum and evaluation standards for school
mathematics. Reston, VA: NCTM.

National Research Council (1989). Ev'erybody counts. Washington, DC: National Academy Press.
Neisser, U. (August, 1989). Direct perception and recognition as distinct perceptual systems. Paper
presented to the Cognitive Science Society, Ann Arbor, MI.
Neuman, D. (1987). The origins of arithmetic skills.: A phenomenographic approach. Goteborg, Swe-
den: Acta Universitates Gothoburgensis.
Newell, A., & Simon, H. A. (1972). Human problem solving. Englewood Cliffs, NJ: Prentice-Hall.
Norman, D. A. (1988). Psychology of Everyday Things. New York, NY: Basic Books.
Pervin, L. A., & Lewis, M. (Eds.) (1978). Perspectives in interactionalpsycholgy. New York, NY: Plenum
Press.

Posner, J., Strike, K., Hewson, P., & Gertzog, W. (1982). Accomodation of a scientific conception:
Toward a theory of conceptual change. Science Education, 66, 211-227.
Purcell, E. M. (1983). The back of the envelope. American Journal of Physics, 51, 11.
Qin, Y., & Simon, H. A. (1990). Imagery and problem solving. The Twelfth Annual Conference of the
Cognitive Science Society (pp. 646-653). Hillsdale, NJ: Lawrence Erlbaum Associates.
Resnick, L. B. (1989). Defining, assessing, and teaching number sense. In J. T. Sowder & B. P.
Schappelle (Eds.), Establishing foundations for research on number sense and related topics: Report
of a conference (pp. 35-39). San Diego, CA: Center for Research in Mathematics and Science
Education, San Diego State University.
Resnick, L. B., & Greeno, J. G. (1990). Conceptual growth of number and quantity. Unpublished
manuscript. University of Pittsburgh: Pittsburgh, PA.
Reys, B. (1989). Conference on number sense: Reflections. In J. T. Sowder & B. P. Schappelle (Eds.),
Establishingfoundationsfor research on number sense and related topics. Report of a conference (pp.
70-73). San Diego, CA: Center for Research in Mathematics and Science Education, San Diego State
University.
Reys, R. E., Rybolt, J. F., Bestgen, B. J., & Wyatt, J. W. (1982). Processes used by good computational
estimators. Journal for Research in Mathematics Education, 13, 183-201.
Riley, M. S., & Greeno, J. G. (1988). Developmental analysis of understanding language about quantities
and of solving problems. Cognition and Instruction, 5, 49-101.
Rogoff, B. (1990). Apprenticeship in thinking. New York, NY: Oxford University Press.
Roschelle, J. (1987). Envisionment, mental models, and physics cognition. Paper presented at the In-
ternational Conference on Artificial Intelligence and Education, Pittsburgh, PA

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
James G. Greeno 217

Roschelle, J., & Greeno, J. G


Berkeley, CA: University of
Saxe, G. B. (1990) Culture and
NJ: Lawrence Erlbaum Associates.

Schegloff, E. A. ( 1981). Discourse as an interactional achievement: Some uses of uh-huh and other thin
that come between sentences. In D. Tannen (Ed.), Analyzing discourse: Text and talk. (pp. 7
Washington, DC: Georgetown University Press.
Schoenfeld, A. H. (1988). Problem solving in context(s). In R. I. Charles & E. A. Silver (Eds.), The tea
and assessing of mathematical problem solving (pp. 82-92). Reston, VA: National Coun
Teachers of Mathematics.

Scribner, S. (1984). Studying working intelligence. In B. Rogoff & J. Lave (Eds.), Everyday cognition:
Its development in social context (pp. 9-40). Cambridge, MA: Harvard University Press.
Shepard, R. N., Kilpatrick, D. W., & Cunningham, J. P. (1975). The internal representation of numbers.
Cognitive Psychology, 7, 82-138.
Silver, E. A. (1989). On making sense of number sense. In J. T. Sowder & B. P. Schappelle (Eds.),
Establishing foundationsfor research on number sense and related topics: Report of a conference (pp.
92-96). San Diego, CA: Center for Research in Mathematics and Science Education, San Diego State
University.
Simon, H. A. (1978). On the forms of mental representation. In C. W. Savage (Ed.), Perception and
cognition: Issues in the foundation ofpsychology (pp. 3-17). Minneapolis, MN: University of Min-
nesota Press.

Smith, C., Suir, J., & Grosslight, L. (1987). Teaching for conceptual change using a computer-based
approach: The case of weight-density differentiation (Report TR87-1 1). Cambridge, MA: Educa-
tional Technology Center, Harvard University.
Steen, L. A. (1988). The science of patterns. Science, 240, 611-616.
Steffe, L. P., & Cobb, P. (1988). Construction of arithmetical meanings and strategies. New York, NY:
Springer-Verlag.
Stucky, S. U. (1989). Situated language use and efficient cognition. Palo Alto, CA: Institute for Research
on Learning.
Suchman, L. (1987). Plans and situated actions. New York, NY: Cambridge University Press.
Thompson, P. (March, 1989). A cognitive model of quantity-based algebraic reasoning. Presented at the
annual meeting of the American Educational Research Association, San Francisco, CA.
Thornton, C. A., & Tucker, S. C. (1989). Lesson planning: The key to developing number sense.
Arithmetic Teacher, 36(6), 18-21.
Toulmin, S. (1972). Human understanding. Princeton, NJ: Princeton University Press.
Trafton, P. R. (1989). Reflections on the number sense conference. In J. R. Sowder & B. P. Schappelle
(Eds.). Establishing foundations for research on number sense and related topics: Report of a
conference (pp. 74-77). San Diego, CA: Center for Research in Mathematics and Science Education,
San Diego State University.
Walk, R. D., & Gibson, E. J. (1961). A comparative and analytical study of visual depth perception.
Psychological Monographs, 75, No. 519.
Warren, W. H., Jr. (1984). Perceiving affordances: Visual guidance of stail climbing. Journal of
Esperimental Psychology: Human Perception and Pelformance, 10, 683-703.
Warren, W. H., Jr., & Whang, S. (1987). Visual guidance of walking through apertures: Body-scaled
information for affordances. Journal of Experimental Psychology: Human Perception and Perfor-
mance, 13, 371-383.

Wellman, H. M., & Estes, D. (1986). Early understanding of mental entities: A reexamination of
childhood realism. Child Development,
Winograd, T., & Flores, F. (1986). Understanding computers and cognition: A new foundation for
design. Norwood, NJ: Ablex.
Wiser, M., & Kipman, D. (1988). The differentiation of heat and temperature:An evaluation of the effect
of microcomputer models on students' misconceptions (Report TR88-20). Cambridge, MA: Educa-
tional Technology Center, Harvard University.

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms
218 Number Sense as Situated Knowing in a Conceptual Domain

Yates, J. (1985). The content of awareness is a model of the world. Psychological Review, 92, 249-284.
Yates, J., Bessman, M., Dunne, M., Jertson, D., Sly, K., & Wendelboe, B. (1988). Are conceptions of
motion based on a naive theory or prototypes? Cognition, 29, 251-275.

AUTHOR

JAMES G. GREENO, Professor of Education, Stanford University; Acting Director, Ins


Research on Learning, Stanford, CA 94305

This content downloaded from 62.204.213.86 on Tue, 05 May 2020 22:22:32 UTC
All use subject to https://about.jstor.org/terms

You might also like