This article was downloaded by: [Northeastern University]On: 26 October 2014, At: 08:35Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH,UK

Journal of the LearningSciencesPublication details, including instructions forauthors and subscription information:http://www.tandfonline.com/loi/hlns20

Reconstructing the LearningSciencesRogers HallPublished online: 17 Nov 2009.

To cite this article: Rogers Hall (2005) Reconstructing the Learning Sciences, Journalof the Learning Sciences, 14:1, 139-155, DOI: 10.1207/s15327809jls1401_8

To link to this article: http://dx.doi.org/10.1207/s15327809jls1401_8

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all theinformation (the “Content”) contained in the publications on our platform.However, Taylor & Francis, our agents, and our licensors make norepresentations or warranties whatsoever as to the accuracy, completeness,or suitability for any purpose of the Content. Any opinions and viewsexpressed in this publication are the opinions and views of the authors, andare not the views of or endorsed by Taylor & Francis. The accuracy of theContent should not be relied upon and should be independently verified withprimary sources of information. Taylor and Francis shall not be liable for anylosses, actions, claims, proceedings, demands, costs, expenses, damages,and other liabilities whatsoever or howsoever caused arising directly orindirectly in connection with, in relation to or arising out of the use of theContent.

This article may be used for research, teaching, and private study purposes.Any substantial or systematic reproduction, redistribution, reselling, loan,

sub-licensing, systematic supply, or distribution in any form to anyone isexpressly forbidden. Terms & Conditions of access and use can be found athttp://www.tandfonline.com/page/terms-and-conditions

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

08:

35 2

6 O

ctob

er 2

014

Reconstructing the Learning Sciences

Jerry A. Fodor, Concepts: Where cognitive science went wrong, Ox-

ford, England: Oxford University Press, 1998, 174 pp., ISBN

0–19–823636–0 (paper).

Geoffrey C. Bowker and Susan Leigh Star, Sorting things out: Clas-

sification and its consequences, Cambridge, MA: MIT Press, 1999,

377 pp., ISBN 0–262–02461–6 (hard).

Commentary by Rogers HallDepartment of Teaching and Learning

Peabody College at Vanderbilt University

“A REAL WITCH always wears a wig to hide her baldness. She wears a first-class

wig. And it is almost impossible to tell a really first-class wig from ordinary hair un-

less you give it a pull to see if it comes off.”

“Then that’s what I’ll have to do,” I said.

“Don’t be foolish,” my grandmother said. “You can’t go ’round pulling at the hair

of every lady you meet, even if she is wearing gloves. Just you try it and see what hap-

pens.”

“So that doesn’t help much either,” I said.

“None of these things is any good on its own,” my grandmother said. “It’s only

when you put them all together that they begin to make a little sense.” (“How to rec-

ognize a witch,” Roald Dahl, 1997, p. 219)

It is hard to locate good, old-fashioned cognition these days. What was once under-

stood by cognitive scientists as the accumulated mental contents of an individual,

shaped by the struggle to solve problems in well-defined task environments, has

given way to a much broader set of approaches in the learning sciences. In this

commentary, I argue that two of these approaches—domain specificity and distrib-

THE JOURNAL OF THE LEARNING SCIENCES, 14(1), 139–155Copyright © 2005, Lawrence Erlbaum Associates, Inc.

Correspondence and requests for reprints should be sent to Rogers Hall, Department of Teaching

and Learning, Peabody College at Vanderbilt University, 162 Wyatt Center, Nashville, TN 37235.

E-mail: r.hall@vanderbilt.edu

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

08:

35 2

6 O

ctob

er 2

014

uted cognition (at least in their strongest forms)—have dramatically altered the ter-

rain for research in the learning sciences. We need new theoretical categories, but,

within this terrain, where should we look?

A LEARNING SCIENCES OF THIN HUMANS?

By arguing for (and empirically demonstrating that) much of the human cognitive

architecture is given to development by the evolution of mind/brain modules in

evolutionary time, domain specificity squeezes ontogenesis (learning by the indi-

vidual) from below. By arguing for (and empirically demonstrating by different

methods that) much of human activity—particularly what we call technical or sci-

entific activity—is enabled by powerful technologies of representation and in-

scription, distributed cognition (and related strands of social constructivism)

squeezes out the individual from above.

What is left in the margin between our biological endowment (as argued) and

the broad sweep of cultural systems? How “thin” has the individual cognitive actor

become, and what are the consequences of this for the learning sciences? The two

books set out for commentary in this column—Jerry Fodor’s (1998) Concepts:

Where Cognitive Science Went Wrong, and Geof Bowker and Leigh Star’s (1999)

Sorting Things Out: Classification and Its Consequences, resonate strongly with

and update our understanding of these two approaches. Although both books are

about concepts and classification as a human activity, they differ radically in their

purposes, methods, and commitments. These differences provide an opportunity to

explore the disintegration of individual cognition on the one hand, and to evaluate

the prospect that advances in the learning sciences may provide a more interesting

“middle ground” for cognitive science on the other.

According to the “domain-specificity” movement in evolutionary psychology

(for a collection of representative and provocative articles, see Hirshfeld &

Gelman, 1994), most of the action in cognition, including what needs to be or can

be learned, has already been determined by constraints on the input–output ports

of brain “modules” that are either sensory or conceptual in nature. These modules

became part of our genetic endowment by providing selective advantage to our an-

cestors as they encountered each other and otherwise challenging environments in

the Pleistocene era. As Cosmides and Tooby (1994) put it, this already endowed ar-

chitecture is where the action is for studies of human cognition:

Many psychologists study the mind without asking what it was designed to do. In-

stead, they hope to uncover its structure by studying things it is capable of doing.

Playing chess, remembering nonsense syllables or long strings of numbers, program-

ming computers, doing college-level statistics—these are all activities that we can

do. It is highly unlikely that the cognitive architecture of the human mind includes

140 HALL

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

08:

35 2

6 O

ctob

er 2

014

procedures that are dedicated to solving any of these problems: The ability to solve

them well would not have enhanced the survival or reproduction of the average Pleis-

tocene hunter-gatherer, and the performance of modern humans on such tasks is gen-

erally poor and uneven. (p. 95)

In sharp contrast, strong versions of actor–network theory and distributed cog-

nition (for central texts, see Latour, 1987, and Hutchins, 1995) hold that individual

cognition is a relatively weak explanatory construct, once we take into account the

historical development and embedding of representational technologies that pro-

vide powerful forms of inscription for the organization of human action. By radi-

cally displacing thinking into cultural artifacts as an explanation for complex hu-

man activity, these approaches appear to squeeze out the theoretical and empirical

relevance of the individual cognitive actor.

In his widely influential comparative analysis of Western and Micronesian nav-

igation, Hutchins (1995) argued

The thinker in this world is a very special medium that can provide coordination

among many structured media—some internal, some external, some embodied in ar-

tifacts, some in ideas, and some in social relationships. (p. 316)

Hutchins’ decentered thinker was celebrated in an enthusiastic review by Latour

(1996), who argued that Hutchins’ analysis revealed a “very lightly equipped hu-

man subject” (p. 56) framed by extensive representational infrastructure. When

analyzed in relation to these technologies, Hutchins’ thinkers become thin enough

for Latour to lift a 10-year moratorium on “cognitive explanations of science and

technology,” originally recommended as a rule of method for science studies in

Science in Action (Latour, 1987, p. 247). Latour (1996) contrasted Hutchins’ thin

humans with their thicker (these are my terms) counterparts from traditional stud-

ies of cognition:

This [Hutchins’ analysis] is the final dissolution of psychology since there is no

agency left that could sustain a psyche at all. Instead of the huge crates and heavy lug-

gage that was necessary before for the internal actor to carry around all the rules and

boxes necessary to think about the world, Hutchins’ thinking agent is more like the

desk of a well-organized executive: empty since everything else has been delegated

outside to something or to someone else. (p. 59)

Hutchins (1996) responded that Latour has gone too far, since work delegated to

technologies or other people stills needs to be analyzed, and these processes of dele-

gationandcoordinationare thecentralproblem forstudiesofdistributedcognition.

Hence my opening remarks. Pushed from below to settle the contents of mind

on evolutionary accounts, one must decide where to draw the line on what is part of

BOOKS & IDEAS 141

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

08:

35 2

6 O

ctob

er 2

014

our endowment and what must be learned (or instructed) to produce a capable (or

even interesting) human actor. Pushed from above to de-center (or distribute) the

individual over networks of “lightly equipped” humans and powerful representa-

tional technologies, one must decide what people need to learn to become partici-

pants in technical or scientific activity (perhaps an easier case than whatever con-

stitutes “everyday” activity) and what to surrender to a sociology of machines. If

one chooses liberal thresholds at both ends, there is very little of the individual left

for the learning sciences to go on about.

If the person as a collection of modules shaped over evolutionary time isn’t a

very appealing object of research in the learning sciences, neither is the person in-

scribed by his or her position in a stable technical or cultural order. What is needed

in the learning sciences is a middle level of structure, a time scale along which cul-

tural arrangements and people change together, in virtue of each other. But finding

this middle level and doing something interesting with it is difficult. Sylvia

Scribner (1985), reviewing the role of history in Vygotsky’s theorizing, called for

just such an elaboration:

Societies and cultural groups participate in world history at different tempos and in

different ways. Each has its own past history influencing the nature of current

change. Particular societies, for example, may adopt the “same” cultural means (e.g.,

writing system) but, as a result of their individual histories, its cognitive implications

may differ widely from one society to the other. (p. 259)

Scribner argued for a level of historical analysis between phylogeny and the lived

history of an individual in society. She called this the “history of individual societ-

ies” and proposed using it to focus on how what appear to be the “same” cultural

artifacts are used and shape people’s thinking in very different ways.

Reviewing this same territory after 15 years of further work in cultural histori-

cal activity theory, Yrjö Engeström (1999) set out a dilemma for designing re-

search that followed learning at both individual and collective levels of analysis:

Historical analysis must be focused on units of manageable size. If the unit is the in-

dividual or the individually constructed situation, history is reduced to ontogeny or

biography. If the unit is the culture or the society, history becomes very general or

endlessly complex. If a collective system is taken as the unit, history may become

manageable, and yet it steps beyond the confines of individual biography. (p. 26)

This problem of locating a productive, historical level of analysis for research in

the learning sciences is, I argue, one path forward in a terrain transformed by do-

main specificity and distributed cognition.

Finding resources for this project is my way of reading these books together. In

overview, I argue that we get very little that is new or useful for the learning sci-

142 HALL

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

08:

35 2

6 O

ctob

er 2

014

ences from Fodor’s proposals for informational atomism (IA). The cognitive sci-

ence Fodor would allow is an architectural project with little to offer research on

learning, teaching, or design—each critical to the learning sciences, as I see the

field. In contrast, Bowker and Star’s social history of classification delivers a wel-

ter of new ideas for analyzing how things (large-scale classification systems, in

particular) and people develop together. Drawing these ideas together for the learn-

ing sciences is my major goal in writing this commentary.

FROM THE GARDEN TO THE LABORATORY

Putting Fodor’s effort into contact with Bowker and Star (and the project of find-

ing new theoretical resources for using history in the learning sciences) takes

some work. This involves distinguishing between DOORKNOBS and WATER,

two criterion cases in Fodor’s argument, then making a bridge across to con-

cepts, categories, and classification as sociotechnical systems. The first two

thirds of Fodor’s book details “where cognitive science went wrong” (from the

cover) regarding what concepts are. Concepts, he argues, cannot be definitions

(philosophy and conceptual analysis have produced no satisfying definitions),

they cannot be prototypes (these don’t compose as concepts should), and they

cannot be abstractions from belief systems (as in “theory theory” accounts of

concepts by analogy to scientific theories). As an alternative, Fodor sets out to

salvage representational theories of mind by proposing a new “doctrine” of IA.

According to IA, “most lexical concepts” have no internal structure, concepts

have semantic content because people’s minds “lock” to relevant statistical regu-

larities (i.e., properties) in the world in a law-like way, and people possess a con-

cept by “being in” this locked relation (p. 121). This new approach to represen-

tational theories of mind is “virgin territory” (p. 121), and Fodor devotes the

remainder of his book to exploring it.

Most relevant for the learning sciences in this exploration is Fodor’s description

of the kinds of concepts he will admit under IA doctrine and how they are related.

Specifically, he allows for:

1. A large collection of primitive concepts that are “mind-dependent” in just

the sense of locking described above (DOORKNOB, RED, and so on.).

2. A collection of “logico-mathematical” concepts that appear to enjoy the

same law-like, locking relation (Fodor does not explore or describe these in

any detail, unfortunately).

3. A set of “natural kind” concepts that people acquire by locking to superfi-

cial appearances that reflect underlying essences (“pretheoretic” versions

of WATER, and so on).

BOOKS & IDEAS 143

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

08:

35 2

6 O

ctob

er 2

014

4. A set of “natural kind concepts as such” that are the hard-won fruit of doing

science to get at essential properties not available through surface similar-

ity (theorized versions of WATER, PROTON, SPIN, and so on).

The developmental story, in brief, is that people possess primitive concepts of type

(1) and (2) by virtue of having experiences that reflect what is stereotypically avail-

able in the world. This is why, Fodor argues, prototype effects are “psychologically

real”without tellinguswhatconceptsare (i.e., theyarenotprototypes, even ifpeople

who possess concepts behave in this way). Statistical regularity is a feature of the

world, and these regularities simply “strike us” when we experience them. Natural

kind concepts of type (3) are acquired because what strikes us at a superficial level is

often correlated with essential but hidden properties. It is only by building and refin-

ing theories that we come to possess concepts of type (4).

So far as the requisite innate endowment is concerned, if the world cooperates you

can get concepts of natural kinds very cheap. That’s what the sticklebacks do; it’s

what Homer did; it’s what children do; it’s what all of us grown-ups do too, most of

the time. By contrast, for you to have a natural kinds concept as such [Type (4)] is for

your link to the essence of the kind not to depend on its inessential properties. This is

a late and sophisticated achievement, historically, ontogenetically, and phylogeneti-

cally, and there is no reason to take it as a paradigm for concept possession at large.

(p. 159, italics in original)

By my reading, natural kind concepts as such are cultural achievements described

by Vygotsky (1986) as “scientific concepts.” But unlike Vygotsky (and many sub-

sequent scholars), Fodor describes these kinds of concepts without offering any

developmental process or mechanism, other than to say that scientific theories play

a mediating role by allowing people to resonate to or lock onto regularities they

would not, without the theory, be able to experience directly. These direct experi-

ences are already structured by human sensory capacities, themselves built on a

collection of more primitive concepts of Types 1 and 2. Evidently, Type 4 concepts

replace Type 3, although again, Fodor does not explore these processes or what

role a cognitive architecture would play in them. He does say that Type 4 concepts

are acquired either by being taught a theory that will enable locking to hidden

properties or by devising and testing theories about these essential properties one-

self. As Fodor puts it, this is how we move from the “garden to the laboratory” (p.

161)—science is the hard work of looking at the world as God does.

You can, if you wish, make a project of getting locked to water in a way that doesn’t

depend on its superficial signs. The easy way to do this is to get some expert to teach

you a theory that expresses the essence of the kind. To be sure, however, that will only

work if the natural kind concept that you’re wanting to acquire is one which some-

144 HALL

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

08:

35 2

6 O

ctob

er 2

014

body else has acquired already. Things get a good deal more difficult if you’re start-

ing ab initio, i.e., without any concepts which express natural kinds as such. (p. 159)

DOORKNOBS OF ALL THINGS!

Two concepts—DOORKNOB and WATER—get extensive treatment by Fodor

(more even than GRANDMOTHER and BACHELOR, which is remarkable for his

variety of philosophical investigation). Fodor uses these examples to redraw the line

between concepts that are “primitive” or “pretheoretic” and those that are available

only through mediating theories that allow people to experience hidden, essential

properties. The former—Types 1, 2, and 3, including DOORKNOB, as Fodor sees

it—are the general case for concepts under IA, and this is what cognitive science

ought to be about. The latter—Type 4 natural kind concepts as such, including

WATERasGodsees it—areaspecial caseof“posttheoretic”scientificachievement.

Analyzing these cultural achievements as though they are the general case—as

though they express regularities about the human mind—is where cognitive science

has gone wrong (e.g., definitional, prototype theory, and theory theory accounts of

concepts). By segregating concepts in this way, Fodor draws a boundary around the

proper concerns of cognitive science (i.e., what those pursuing representational the-

ories of mind should do to understand the human cognitive architecture).

Although it is easy to understand how WATER (its constituents HYDROGEN

and OXYGEN, their relations, and so on) would be a cultural achievement involving

scientific theory, it is not so easy to see DOORKNOB as a primitive concept, innate

and without internal structure. This is exactly why Fodor analyzes “DOORKNOB,

of all things!” (p. 123) as an example of how concepts most typically work. It is a

monster for the nativism–empiricism controversy in cognitive science, as tradition-

allyconceived (Keil, 2000). According to Fodor, humans lock to doorknobs in a mu-

tual relation between human physical abilities (grasping, pulling, and so on) and the

surfacepropertiesof suchobjects (their roundness, connectivity todoors, andsoon).

All this happens without needing to represent the internal structure of the concept or

to worry about what role it plays in other schemes of inference. People’s minds and

DOORKNOB just bind together in the mundane act of getting in and out of build-

ings, and there is no need for a cognitive or psychological theory of learning to ex-

plain it. Moreover, this is the way it works for most concepts.

In contrast with Scribner’s (1985) call for a “history of individual societies” as a

productive level of comparative analysis, Fodor’s IA is a remarkable simplification

of the terrain for cognitive science, and, I argue, much too narrow a framing to

make for an interesting (or even somewhat wide-ranging) learning sciences. To see

this problem clearly, we must contrast Fodor’s analysis of DOORKNOB with

Bowker and Star’s analysis of infrastructure. To get started, I take a short detour

through a provocative (but seldom cited) essay by Johnson and Latour (1995), in

BOOKS & IDEAS 145

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

08:

35 2

6 O

ctob

er 2

014

which the now largely-invisible problem of allowing humans to pass through

walled enclosures is unpacked into a chain of technical developments.

In Fodor’s way of thinking, doorknobs are instances of DOORKNOB in just the

sense sketched previously, with all the relevant distinctions given to development

on an evolutionary time scale. But for Johnson and Latour (1995), doorknob is a

sociotechnical system that involves a complex series of displacements that bind in-

dividuals, society, and the material world together. Through an extended thought

experiment, Johnson and Latour obtained a list of eight developments regarding

the way to produce what Fodor treats as an atomic chunk of information. Each de-

velopment solves a problem by shifting capacities between humans and machines.

The list is as follows. Johnson and Latour’s terms are in italics, and following each

is a brief description of an old problem solved and a new problem posed, as the de-

velopment progresses:

The wall … a container, but what is in cannot get out, and what is out cannot

get in

Holes in the wall … essential for entry/egress, but must be broken open

and repaired on each use

A hole-wall (or door) … eliminates breaking and repairing, but requires

lifting and replacing a heavy door

The hinge-pin … incorporates a lever to ease lifting, but users leave the

door open

At this point in the developmental sequence, Johnson and Latour pause to summa-

rize their method:

To size up the work done by hinges, you simply have to imagine every time you want

to get in or out of the building you have to do the same work as a prisoner trying to es-

cape or a gangster trying to rob a bank, plus the work of those who rebuild either the

prison’s or the bank’s walls. (p. 258)

Disciplined users … highly trained, moral folk will close the door, but peo-

ple as a class are notoriously unreliable

A porter (French for “door”) … only one concierge must be disciplined,

but they may resist or go on strike

A groom or spring … the door closes, but the spring resembles a rude por-

ter, slamming the door in your face

Instructions … users can be trained, but they again are too many and un-

reliable

A hydraulic door closer … energy is collected on entry, and then smoothly

used to close the door

146 HALL

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

08:

35 2

6 O

ctob

er 2

014

Johnson and Latour’s contribution to the concept of DOORKNOB calls atten-

tion to how agency is delegated into technical devices in an exchange between peo-

ple and machines over what Scribner (1985) called societal history. Although the

biological endowment of human participants (Fodor’s sense of history) may be im-

portant for what develops or needs to be learned, we get little purchase on the qual-

ity of what a doorknob is, how it has come to have this quality as a cultural artifact,

or what role humans play in making or using it in consideration of that biological

endowment. In saying this, I do not mean to argue that constraints and affordances

associated with human biology have nothing to do with good or poor design, just

that they tell us relatively little about the development or even the success of de-

signed artifacts. Johnson and Latour’s door-closer is a dramatically more interest-

ing entity for the learning sciences than is Fodor’s DOORKNOB, and for just the

set of reasons that Bowker and Star explore in their book.

We need analytic categories that open up new perspectives on this mid-level his-

tory—not subsumed by the history of the species (whatever has happened along an

evolutionary time scale that gives us brain modules) and not lost to historical analysis

in the moment, during fragments of interaction in which the individual might be said to

have learned something through solving a problem. The most interesting questions in

the learning sciences are those that spread out from the middle in productive ways.

A SOCIAL HISTORY OF LARGE-SCALECLASSIFICATION SYSTEMS

Bowker and Star give us a substantive expansion of Latour’s approach to the social

construction (and meaning) of technical and scientific knowledge and artifacts.

Like Latour, they give symmetric treatment to things that people make, but they in-

clude less visible and more widespread artifacts that congeal human labor in physi-

cal or structural form. In particular, they focus on large-scale classification sys-

tems. According to Bowker and Star’s analysis, these systems have a contested

history of production, they become invisible (or naturalized) in current use, and

they shape human thinking and action in ways that are fundamental to understand-

ing learning and development.

For the learning sciences research community, there is no time like the present

for a social history of large-scale classification systems. For example, educational

researchers in the United States are in the grip of a rush towards“high standards”

and “high stakes assessments” in public schooling, so that “no child be left be-

hind.”1 Under these circumstances, we could use a critical analysis of how

BOOKS & IDEAS 147

1See the earlier Books & Ideas article on “Reassessing School Standards” (The Journal of the

Learning Sciences, 13[2]) for a more extensive discussion of this particular large-scale classification

system.

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

08:

35 2

6 O

ctob

er 2

014

policymakers and information system designers go about cleaning up or reforming

other people’s activities in different areas of social life. Just how would one create

explicit standards for teaching and learning, on the one hand, then on the other de-

velop technologies that classify the to-be-reformed activities of teachers and learn-

ers as meeting or (more consequentially) failing to meet these standards?

Through comparative analysis, Bowker and Star show us both how large-scale

classification systems become powerful and how their capacity for control produces

suffering for people who are either directly injured bybeing attached to a categoryor

who come to inhabit the borderlands created by the competing demands of different

classification systems. In this sense, the book provides new ideas for analyzing how

fitting poorly into classification systems can have consequences for people, but also

how responses or resistance to a lack of fit can create the grounds for developing new

forms of classification that may produce less suffering.

The analysis proceeds through a sequence of four case studies: cause of death

classification in the International Classification of Diseases (ICD), the experience

of time among those diagnosed with tuberculosis (TB) and confined to institu-

tional care, race classification on the basis of skin color (White, Colored, or Black)

in apartheid South Africa, and classification of nursing activities as professionally

distinct and billable services. Presented in this order, the cases take the reader from

aspects of classification that are distant in time and have a large-scale spatial distri-

bution (the ICD has been in continuous use, and revision, since the 1890s), into as-

pects of classification that are deeply consequential (and contested) for the individ-

ual being attached to a category (a diagnosis of TB or a classification of one’s skin

color as a biographical event), and finally to the prospective problem of designing

a system that sorts things out without undue suffering (classification of profes-

sional services in work practice).

METHODS FOR INVERTING INFRASTRUCTURE

Bowker and Star set out to study classification systems that are so ubiquitous, so

deeply “sunk” in the historical sedimentation of everyday life, that they have be-

come invisible. Making these systems visible presents a variety of methodological

problems. They start out by pointing out that every actual system they have studied

would fail to meet a classical definition (i.e., what they call “Aristotelian”) using

criteria of unique classificatory principles, mutually exclusive categories, and

complete description. Setting aside this a priori definition (recall that Fodor also

argues that classical accounts of concepts fail), they instead take a pragmatist

stance, arguing “anything consistently called a classification system and treated as

such can be included in the term” (p. 13, italics in original). By bracketing the “true

nature” of classification systems, Bowker and Star hope to open an investigation of

the work people do to negotiate exactly these tensions between formal definitions

148 HALL

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

08:

35 2

6 O

ctob

er 2

014

and actual practice. Classification systems can be described in classical terms, and

these descriptions may be a necessary part of their development and use, but the

work entailed is always (and productively) messier than the formal description.

This is the work through which people and things are attached to categories and the

category systems themselves are developed and made stable.

Just what constitutes actual practices of classification is complicated. Bowker

and Star make a series of framing assumptions, which I have assembled into the

following list. These assumptions help realize the title of their book. “Sorting

things out” is the practical work of classification, either in use or under design.

1. Classification systems arise within communities of practice (Lave &

Wenger, 1991; Wenger, 1999), and these can be analyzed as social worlds.

2. Every human moment involves the work of “doing being ordinary” (Sacks,

1992) with respect to these systems, and this work is “neither created nor

destroyed, yet may be radically reshaped” (Bowker & Star, 1999, p. 231)

when a system changes.

3. Some classification systems are naturalized across so many communities

that they appear invisible or inevitable as standards (e.g., mathematics, ra-

tional models of planning).

4. There is a huge power differential between makers (producers) and users

(consumers) of large-scale classification systems, and this leads to predict-

able (and resistible) kinds of suffering.

5. Most important for work in the learning sciences, there is no “great divide”

(following Latour, 1987) between local–global or folk–scientific classifi-

cation systems. Instead, the scale or precision of classification systems re-

flects generative tensions, and these require a developmental analysis.

Two other methodological strategies are notable in the book. First, Bowker and

Star combine historical analysis of classification systems—how they are devel-

oped, maintained, and eventually retired (they give examples of “category death”

over time)—with an analysis of the biographical trajectories of people and things

being classified by these systems. For example, types of TB are classified by a

changing array of operational tests, and these tests impose a metric over the pa-

tient’s experience of time, body, and self. What results is a body-biography trajec-

tory in which the identity of patients is “torqued” as they become incumbents of

medical, diagnostic categories. Because TB is a progressive disease for some, but

can become “inactive” for others, institutional life for TB patients becomes an on-

going struggle to locate oneself within this larger classificatory “timetable.” Days

are filled with the details of measurements relevant to the classification, and one’s

institutional stay becomes a kind of “time out” from other life activities—that is, a

significant biographical event for one’s sense of self. This combination of history

and biography helps us to see the horizon along which classification systems shape

BOOKS & IDEAS 149

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

08:

35 2

6 O

ctob

er 2

014

personal identity (biographies of attachment) even as these systems are stabilized

or changed in the process.

Second, Bowker and Star analyze the narrative resources provided in bureau-

cratic forms for “telling stories” about people and things being classified. This se-

lective provision, they argue, shapes the kinds of biographies or identities that are

possible through a process of “convergence” in which the world and the system for

describing it develop together (i.e., not a map of the territory but a map in the terri-

tory). For example, the ICD cuts the human life cycle into time intervals that are

highly selective. For infants and children—in which many complications are pos-

sible—human time is finely divided. In contrast, being an “adult” is comparatively

timeless, crowded into a single time interval (15 to 124 years) without a well-de-

fined end. Although it was once possible to die of “old age,” the narrative resources

for this category have dwindled over the past century (55 categories related to old

age in the 1913 ICD, but only 4 by 1992). As the classification system and experi-

ence of the life cycle converge over time, Bowker and Star argue, the individual be-

comes a “tabula rasa onto which various diseases are inscribed” (p. 90) using the

narrative resources of the bureaucratic form.

FORGETTING AND THE DEVELOPMENT OFPEOPLE AND THINGS

I find two aspects of Bowker and Star’s analysis particularly important for a theory

of learning and development that articulates between individual and collective ac-

tion. The first aspect concerns how classification systems, under design and in use,

operate on a distinction between remembering and forgetting. Drawing from their

participation in and study of the design of a new information system for classifying

nursing work, Bowker and Star analyze the role of forgetting in the design of clas-

sification systems. Nursing practitioners need to find a classification that will

make their work visible (and valuable), without surrendering control of that work

to other, more powerful professional groups (e.g., doctors, academics, or insurance

providers). Under these circumstances (the usual organizational context of de-

sign), what the classification system insists you “forget”—by eliminating narrative

resources for formal description and so making things invisible—is just as impor-

tant as what it makes you “remember”—by providing specific concepts and requir-

ing full description in these terms. As Bowker and Star put it,

To deal with the plenum of information that all good organizations logically need,

one can operate a distribution of memory in space (such and such a subgroup needs to

hold such and such knowledge) and a distribution of memory in time (such and such a

memory will only be recalled if a given occasion arises). Classification systems pro-

150 HALL

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

08:

35 2

6 O

ctob

er 2

014

vide both a warrant and a tool for forgetting at the same time as they operate this dis-

tribution. (pp. 276–277)

Classification systems provide selective visibility; this is a double-edged sword for

those being attached to concepts and categories within the system, and these ten-

sions are the focus of concerted attention during design.

For an example more related to educational research, a recent study by Ilana

Horn (2002) explored this tension in the organizational context of two re-

form-oriented, public high school mathematics departments. Both are under dis-

trict pressure to eliminate remedial mathematics courses, which disproportion-

ately “acquire” (McDermott, 1993) ethnic minority students and make it

unlikely that they will leave high school with a transcript acceptable to college

admissions officers. In a department in which equity-driven reforms are seen as

a matter of school-wide culture; mathematics teachers respond by sorting fixed

categories of students—described by their moral and intellectual qualities as

“college bound,” “lazy,” or “screwups”—into a renamed set of courses, none of

which would be recognizable by district officials as “remedial” (e.g., for

screwups, an introductory algebra course is separated into two, semester-length

courses that can be taken an indefinite number of times). In a department in

which equity-driven reforms are pursued in subject-matter departments, mathe-

matics teachers describe students in less essential terms as “fast” and “slow,”

then focus on whether their teaching in college-prep classrooms is accessible to

both. The second department, according to Horn’s analysis, produces much

better achievement among ethnic minority youth, but at great personal cost to

participating teachers. In the first department, learners fare less well, but teach-

ers are able to retain control over their work (e.g., more experienced teachers

teach college-prep sections). In both cases, Horn shows us how departmental

meetings are sites for working out the relations between formal and informal

classification systems in ways that make the work of different participants

(teachers and their students) more or less visible, and consequentially so.

A second major contribution of Bowker and Star’s book is to link together

membership (of people) and naturalization (of things) in a process of convergence

that describes how people-and-things develop, together. This interleaving of peo-

ple and things expands social practice theories of learning. Bowker and Star start

with the idea that legitimate peripheral participation describes a trajectory of mem-

bership as a person enters some community of practice (Lave & Wenger, 1991),

but they also point out that these relations of membership are almost always medi-

ated by objects or things (technical devices, categories, stories, and so on) that are

themselves undergoing development. They describe this as the “trajectory of natu-

ralization” of an object as used in the practices of a community (i.e., moving from

strange, to familiar, to invisible for members). These two trajectories are inter-

leaved and pull at each other in a process of convergence.

BOOKS & IDEAS 151

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

08:

35 2

6 O

ctob

er 2

014

It is possible for analytical purposes to think of two trajectories traveling in tandem,

membership and naturalization. Just as it is not practically possible to separate a dis-

ease from a sick patient, yet it is possible to speak of the trajectories of disease and bi-

ography operating and pulling at one another, as seen … in the case of tuberculosis.

(Bowker & Star, 1999, p. 300)

When analysts arrive late in this cycle of convergence, they find people doing

things that are taken for granted in the community. In this sense, cognition is al-

ready distributed over things and people, often in complex ways. Johnson and

Latour’s (1995) de-inscription of a door-closer pulls this complexity apart for a

technical artifact, but what about a classification that seems more primitive, like

judging color (e.g., RED is unproblematically of Type 1 for Fodor)? Chuck

Goodwin (1994) provided a careful analysis of the “structure of intentionality”

built into the Munsell color chart, as used by archaeologists who need to judge the

color of dirt they excavate to identify architectural remains that make up data for

their studies. Talk and action in the pit are oriented to fixing the quality of dirt

against standard color samples, and this is made possible by the systematic juxta-

position of color, symbolic codes, and holes punched in the Munsell chart. Bits of

dirt, either still in the ground or on a trowel, are positioned under holes in the pub-

lished chart until a good visual match is found and its symbolic description re-

corded on a data sheet. These coordinated bits of action, talk, and inscription go

from dirt to symbol in what appears to be a relatively simple transformation (see

also Latour, 1995), but this is a transformation made simple only by the prior con-

ventions of the chart, the widespread adoption of these conventions in archaeologi-

cal practice (otherwise the symbolic residue is not “data,” because it cannot be

combined with others’ work), and the mundane (though necessary and ongoing)

work of disciplining the perception of fieldworkers so they can reliably make these

visual discriminations (see also Stevens & Hall, 1998). It is in this sense that the in-

tentions of fieldworkers—what they mean and do when attaching a bit of dirt to a

color category—is structured by prior history. Their actions are shaped by and

have meaning only in terms of that prior history—this is what convergence makes

possible through (returning to Scribner, 1985) the “history of individual societies.”

When analysts arrive during the cycle of convergence, either when existing infra-

structure breaks down or when people explicitly set out to build new technical ar-

rangements for having and using concepts, people are often busy with the work of

distributing cognition. For example, in a comparative analysis of consulting meet-

ings in field entomology and architectural design (Hall, Stevens, & Torralba, 2002),

wefoundstudyparticipantsbuildingnewrepresentationaldevices toclassifyinsects

or arguing against widely adopted building codes to classify historic buildings as

safe forpublicuse. Ineachcase,peoplecriticizedexisting technologiesofclassifica-

tion, consideredalternativemeansof makingcategoryjudgments (e.g., different ter-

mite species, safe or unsafe buildings), and looked forward to work they would need

152 HALL

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

08:

35 2

6 O

ctob

er 2

014

to do if they were to adopt one of the alternatives under consideration. There were

striking differences in the ways that specialists from different disciplines attended to

andusedshared representationsof the“same”objects.Forexample, lookingat apair

of graphs showing an abundance of chemicals stripped from the exoskeletons of ter-

mite samples, field entomologists and chemists oriented to the relative size of indi-

vidual peaks (each representing a particular hydrocarbon) and the geographic loca-

tion from which the samples were taken. But a statistician, looking at the same two

graphs of abundance, described and acted on them as two points in a common metric

space. In the ensuing conversation, finding a new way to classify same and different

termites required preserving both ways of looking in a manner that would make re-

sulting claims about new termite species accountable to a wider audience of ento-

mologists, biochemists, and research sponsors.

TAKING STOCK

Looking back over these two books, it is difficult to imagine some combination of

their proposals that would lead to a livelier learning sciences. If Fodor gets to cordon

off even more (“most,” as he puts it) of human cognition inside a collection of innate

modules, and if “psychology [is] swept inside out” by distributed cognition (as

Latour would have it), then we are in for a major contraction of the field. But as I have

argued in this commentary, there is more to concepts such as DOORKNOB or RED

thanFodorwouldallow,andit isentirelypossible tostudyhowtheactionsof individ-

uals contribute to (or resist and undermine) the power of classification systems.

It may be helpful to point out that Bowker and Star, though not setting out to re-

construct cognitive science, do help us understand how Fodor is going about his re-

construction. Fodor proposes for cognitive science what Bowker and Star would call

a “clearance” (p. 258), an attempt to erase the prior history of the field and start over

with a proper science. Fodor’s is a proposal for wholesale forgetting—of classical

definitions, prototype theory, theory theory, and empiricism more generally—to be

replaced with a new doctrine of IA. Although some concepts are learned or taught,

most (and thosemost important forunderstandinghumans)arenot,hewouldhaveus

believe. As I have argued throughout this commentary, I do not find this a promising

path, particularly not for something called the learning sciences.

This is what makes Bowker and Star’s contributions so valuable, in my view, par-

ticularly to researchers interested in attending simultaneously to individual and col-

lective processes of learning and development. By broadening our focus from con-

cepts to large-scale classification systems, we can hold the individual and collective

in view at the same time. Concepts are part of larger sociotechnical systems, and al-

thoughthesesystemsmaybetunedtoourbiologicaldispositions (what“strikes”us),

they are not determined (or constructed) by them. Instead, classification systems

(and hence many important concepts, in actual use) are constructed over relatively

BOOKS & IDEAS 153

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

08:

35 2

6 O

ctob

er 2

014

recent historical time, through processes that can be studied as a form of human

work. We simply need designs in the learning sciences that encompass different

timescales and different levels of analysis, a project that is already underway (Cole,

1996; Engeström, 1999; Hall et al., 2002; Lemke, 2002; Saxe, 1991). Bowker and

Star’s contributions—methods for inverting infrastructure, innovation, and resis-

tance in thegapsbetweenclassificationsystems; torquedbiographiesof categoryat-

tachment; and intertwined trajectories of membership and naturalization during cy-

cles of convergence—move us further along in this project.

ACKNOWLEDGMENTS

My thanks to members of the Representational Practices group at the Univer-

sity of California–Berkeley—Flavio Azevedo, Coe Leta Finke, Dan Glaser, Bruce

Goldstein, Charles Hammond, Lani Horn, Gwen Ottinger, Tamar Posner, Ann

Ryu, and Chris Wu—for a lively reading of Bowker and Star. I faced Fodor alone.

Thanks also to Karen Wieckert for a careful reading at the end, and to Maren and

Kela for extended discussions of Roald Dahl’s proposals.

REFERENCES

Cole, M. (1996). Cultural psychology: A once and future discipline. Cambridge, MA: Harvard Univer-

sity Press.

Cosmides, L., & Tooby, J. (1994). Origins of domain specificity: The evolution of functional organiza-

tion. In L. A. Hirschfeld & S. A. Gelman (Eds.), Mapping the mind: Domain specificity in cognition

and culture (pp. 85–116). Cambridge, England: Cambridge University Press.

Dahl, R. (1997). The Roald Dahl treasury. New York: Viking.

Engeström, Y. (1999). Activity theory and individual and social transformation. In Y. Engeström, R.

Miettinen, & U. Punamaki (Eds.), Perspectives on activity theory (p. 26). Cambridge, England: Cam-

bridge University Press.

Goodwin, C. (1994). Professional vision. American Anthropologist, 96, 606–633.

Hall, R., Stevens, R., & Torralba, A. (2002). Disrupting representational infrastructure in conversations

across disciplines. Mind, Culture, & Activity, 9, 179–210.

Hirschfeld, L. A., & Gelman, S. A. (1994). Mapping the mind: Domain specificity in cognition and cul-

ture. New York: Cambridge University Press.

Horn, I. S. (2002). Learning on the job: Mathematics teachers’ professional development in the con-

texts of high school reform. Unpublished doctoral dissertation, University of California, Berkeley.

Hutchins, E. (1995). Cognition in the wild. Cambridge, MA: MIT Press.

Hutchins, E. (1996). Response to reviewers. Mind, Culture, & Activity, 3, 64–68.

Johnson, J., & Latour, B. (1995). Mixing humans and nonhumans together: The sociology of a

door-closer. In S. L. Star (Ed.), Ecologies of knowledge: Work and politics in science and technology

(pp. 257–277). Albany: State University of New York Press.

Keil, F. C. (2000). Nurturing nativism, review of Cowie’s What’s within. In M. Nani & M. Marraffa

(Eds.), A field guide to the philosophy of mind. Retrieved November 2, 2004, from the World Wide

Web: http://www.uniroma3.it/kant/field/cowiesymp.htm).

154 HALL

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

08:

35 2

6 O

ctob

er 2

014

Latour, B. (1987). Science in action: How to follow scientists and engineers through society. Cam-

bridge, MA: Harvard University Press.

Latour, B. (1995). The “Pedofil” of Boa vista: A photo-philosophical montage. Common Knowledge, 4,

144–187.

Latour, B. (1996). Cogito ergo sumus! Or psychology swept inside out by the fresh air of the upper

deck. Review symposium: Cognition in the wild. Mind, Culture, & Activity, 3, 54–63.

Lave, J., & Wenger, E. (1991). Situated learning: Legitimate peripheral participation. Cambridge:

Cambridge University Press.

Lemke, J. L. (2002). The long and short of it: Comments on multiple timescale studies of human activ-

ity. The Journal of the Learning Sciences, 10, 17–26.

McDermott, R. (1993). The acquisition of a child by a learning disability. In S. Chaiklin & J. Lave

(Eds.), Understanding practice (pp. 269–305). New York: Cambridge University Press.

Sacks, H. (1992). Doing “being ordinary.” In G. Jefferson (Ed.), Lectures on conversation (Vol. II, pp.

215–221). Oxford, England: Blackwell.

Saxe, G. B. (1991). Culture and cognitive development. Hillsdale, NJ: Lawrence Erlbaum Associates,

Inc.

Scribner, S. (1985). Vygotsky’s uses of history. In J. V. Wertsch (Ed.), Culture, communication and

cognition (pp. 119–145). Cambridge, England: Cambridge University Press.

Stevens, R., & Hall, R. (1998). Disciplined perception: Learning to see in technoscience. In M. Lampert

& M. Blunk (Eds.), Talking mathematics in school: Studies of teaching and learning (pp. 107–149).

Cambridge, England: Cambridge University Press.

Vygotsky, L. (1986). Thought and language. Cambridge, MA: MIT Press.

Wenger, E. (1999). Communities of practice: Learning, meaning and identity. Cambridge: Cambridge

University Press.

BOOKS & IDEAS 155

Dow

nloa

ded

by [

Nor

thea

ster

n U

nive

rsity

] at

08:

35 2

6 O

ctob

er 2

014