Saturday, October 15, 2005

What is the best program?

"To be effective, every knowledge worker, and especially every executive, therefore needs to be able to dispose of time in fairly large chunks. To have small dribs and drabs of time at his disposal will not be sufficient even if the total is an impressive number of hours.

This is particularly true with respect to time spent working with people, which is, of course, a central task in the work of the executive. People are time-consumers. And most people are time-wasters."

-- Peter Drucker, The Effective Executive, 1966.


October 16, 2005
Meet the Life Hackers


In 2000, Gloria Mark was hired as a professor at the University of California at Irvine. Until then, she was working as a researcher, living a life of comparative peace. She would spend her days in her lab, enjoying the sense of serene focus that comes from immersing yourself for hours at a time in a single project. But when her faculty job began, that all ended. Mark would arrive at her desk in the morning, full of energy and ready to tackle her to-do list - only to suffer an endless stream of interruptions. No sooner had she started one task than a colleague would e-mail her with an urgent request; when she went to work on that, the phone would ring. At the end of the day, she had been so constantly distracted that she would have accomplished only a fraction of what she set out to do. "Madness," she thought. "I'm trying to do 30 things at once."

Lots of people complain that office multitasking drives them nuts. But Mark is a scientist of "human-computer interactions" who studies how high-tech devices affect our behavior, so she was able to do more than complain: she set out to measure precisely how nuts we've all become. Beginning in 2004, she persuaded two West Coast high-tech firms to let her study their cubicle dwellers as they surfed the chaos of modern office life. One of her grad students, Victor Gonzalez, sat looking over the shoulder of various employees all day long, for a total of more than 1,000 hours. He noted how many times the employees were interrupted and how long each employee was able to work on any individual task.

When Mark crunched the data, a picture of 21st-century office work emerged that was, she says, "far worse than I could ever have imagined." Each employee spent only 11 minutes on any given project before being interrupted and whisked off to do something else. What's more, each 11-minute project was itself fragmented into even shorter three-minute tasks, like answering e-mail messages, reading a Web page or working on a spreadsheet. And each time a worker was distracted from a task, it would take, on average, 25 minutes to return to that task. To perform an office job today, it seems, your attention must skip like a stone across water all day long, touching down only periodically.

Yet while interruptions are annoying, Mark's study also revealed their flip side: they are often crucial to office work. Sure, the high-tech workers grumbled and moaned about disruptions, and they all claimed that they preferred to work in long, luxurious stretches. But they grudgingly admitted that many of their daily distractions were essential to their jobs. When someone forwards you an urgent e-mail message, it's often something you really do need to see; if a cellphone call breaks through while you're desperately trying to solve a problem, it might be the call that saves your hide. In the language of computer sociology, our jobs today are "interrupt driven." Distractions are not just a plague on our work - sometimes they are our work. To be cut off from other workers is to be cut off from everything.

For a small cadre of computer engineers and academics, this realization has begun to raise an enticing possibility: perhaps we can find an ideal middle ground. If high-tech work distractions are inevitable, then maybe we can re-engineer them so we receive all of their benefits but few of their downsides. Is there such a thing as a perfect interruption?

Mary Czerwinski first confronted this question while working, oddly enough, in outer space. She is one of the world's leading experts in interruption science, and she was hired in 1989 by Lockheed to help NASA design the information systems for the International Space Station. NASA had a problem: how do you deliver an interruption to a busy astronaut? On the space station, astronauts must attend to dozens of experiments while also monitoring the station's warning systems for potentially fatal mechanical errors. NASA wanted to ensure that its warnings were perfectly tuned to the human attention span: if a warning was too distracting, it could throw off the astronauts and cause them to mess up million-dollar experiments. But if the warnings were too subtle and unobtrusive, they might go unnoticed, which would be even worse. The NASA engineers needed something that would split the difference.

Czerwinski noticed that all the information the astronauts received came to them as plain text and numbers. She began experimenting with different types of interruptions and found that it was the style of delivery that was crucial. Hit an astronaut with a textual interruption, and he was likely to ignore it, because it would simply fade into the text-filled screens he was already staring at. Blast a horn and he would definitely notice it - but at the cost of jangling his nerves. Czerwinski proposed a third way: a visual graphic, like a pentagram whose sides changed color based on the type of problem at hand, a solution different enough from the screens of text to break through the clutter.

The science of interruptions began more than 100 years ago, with the emergence of telegraph operators - the first high-stress, time-sensitive information-technology jobs. Psychologists discovered that if someone spoke to a telegraph operator while he was keying a message, the operator was more likely to make errors; his cognition was scrambled by mentally "switching channels." Later, psychologists determined that whenever workers needed to focus on a job that required the monitoring of data, presentation was all-important. Using this knowledge, cockpits for fighter pilots were meticulously planned so that each dial and meter could be read at a glance.

Still, such issues seemed remote from the lives of everyday workers - even information workers - simply because everyday work did not require parsing screenfuls of information. In the 90's, this began to change, and change quickly. As they became ubiquitous in the workplace, computers, which had until then been little more than glorified word-processors and calculators, began to experience a rapid increase in speed and power. "Multitasking" was born; instead of simply working on one program for hours at a time, a computer user could work on several different ones simultaneously. Corporations seized on this as a way to squeeze more productivity out of each worker, and technology companies like Microsoft obliged them by transforming the computer into a hub for every conceivable office task, and laying on the available information with a trowel. The Internet accelerated this trend even further, since it turned the computer from a sealed box into our primary tool for communication. As a result, office denizens now stare at computer screens of mind-boggling complexity, as they juggle messages, text documents, PowerPoint presentations, spreadsheets and Web browsers all at once. In the modern office we are all fighter pilots.

Information is no longer a scarce resource - attention is. David Rose, a Cambridge, Mass.-based expert on computer interfaces, likes to point out that 20 years ago, an office worker had only two types of communication technology: a phone, which required an instant answer, and postal mail, which took days. "Now we have dozens of possibilities between those poles," Rose says. How fast are you supposed to reply to an e-mail message? Or an instant message? Computer-based interruptions fall into a sort of Heisenbergian uncertainty trap: it is difficult to know whether an e-mail message is worth interrupting your work for unless you open and read it - at which point you have, of course, interrupted yourself. Our software tools were essentially designed to compete with one another for our attention, like needy toddlers.

The upshot is something that Linda Stone, a software executive who has worked for both Apple and Microsoft, calls "continuous partial attention": we are so busy keeping tabs on everything that we never focus on anything. This can actually be a positive feeling, inasmuch as the constant pinging makes us feel needed and desired. The reason many interruptions seem impossible to ignore is that they are about relationships - someone, or something, is calling out to us. It is why we have such complex emotions about the chaos of the modern office, feeling alternately drained by its demands and exhilarated when we successfully surf the flood.

"It makes us feel alive," Stone says. "It's what makes us feel important. We just want to connect, connect, connect. But what happens when you take that to the extreme? You get overconnected." Sanity lies on the path down the center - if only there was some way to find it.

It is this middle path that Czerwinski and her generation of computer scientists are now trying to divine. When I first met her in the corridors of Microsoft, she struck me as a strange person to be studying the art of focusing, because she seemed almost attention-deficit disordered herself: a 44-year-old with a pageboy haircut and the electric body language of a teenager. "I'm such a spaz," she said, as we went bounding down the hallways to the cafeteria for a "bio-break." When she ushered me into her office, it was a perfect Exhibit A of the go-go computer-driven life: she had not one but three enormous computer screens, festooned with perhaps 30 open windows - a bunch of e-mail messages, several instant messages and dozens of Web pages. Czerwinski says she regards 20 solid minutes of uninterrupted work as a major triumph; often she'll stay in her office for hours after work, crunching data, since that's the only time her outside distractions wane.

In 1997, Microsoft recruited Czerwinski to join Microsoft Research Labs, a special division of the firm where she and other eggheads would be allowed to conduct basic research into how computers affect human behavior. Czerwinski discovered that the computer industry was still strangely ignorant of how people really used their computers. Microsoft had sold tens of millions of copies of its software but had never closely studied its users' rhythms of work and interruption. How long did they linger on a single document? What interrupted them while they were working, and why?

To figure this out, she took a handful of volunteers and installed software on their computers that would virtually shadow them all day long, recording every mouse click. She discovered that computer users were as restless as hummingbirds. On average, they juggled eight different windows at the same time - a few e-mail messages, maybe a Web page or two and a PowerPoint document. More astonishing, they would spend barely 20 seconds looking at one window before flipping to another.

Why the constant shifting? In part it was because of the basic way that today's computers are laid out. A computer screen offers very little visual real estate. It is like working at a desk so small that you can look at only a single sheet of paper at a time. A Microsoft Word document can cover almost an entire screen. Once you begin multitasking, a computer desktop very quickly becomes buried in detritus.

This is part of the reason that, when someone is interrupted, it takes 25 minutes to cycle back to the original task. Once their work becomes buried beneath a screenful of interruptions, office workers appear to literally forget what task they were originally pursuing. We do not like to think we are this flighty: we might expect that if we are, say, busily filling out some forms and are suddenly distracted by a phone call, we would quickly return to finish the job. But we don't. Researchers find that 40 percent of the time, workers wander off in a new direction when an interruption ends, distracted by the technological equivalent of shiny objects. The central danger of interruptions, Czerwinski realized, is not really the interruption at all. It is the havoc they wreak with our short-term memory: What the heck was I just doing?

When Gloria Mark and Mary Czerwinski, working separately, looked at the desks of the people they were studying, they each noticed the same thing: Post-it notes. Workers would scrawl hieroglyphic reminders of the tasks they were supposed to be working on ("Test PB patch DAN's PC - Waiting for AL," was one that Mark found). Then they would place them directly in their fields of vision, often in a halo around the edge of their computer screens. The Post-it notes were, in essence, a jury-rigged memory device, intended to rescue users from those moments of mental wandering.

For Mark and Czerwinski, these piecemeal efforts at coping pointed to ways that our high-tech tools could be engineered to be less distracting. When Czerwinski walked around the Microsoft campus, she noticed that many people had attached two or three monitors to their computers. They placed their applications on different screens - the e-mail far off on the right side, a Web browser on the left and their main work project right in the middle - so that each application was "glanceable." When the ding on their e-mail program went off, they could quickly peek over at their in-boxes to see what had arrived.

The workers swore that this arrangement made them feel calmer. But did more screen area actually help with cognition? To find out, Czerwinski's team conducted another experiment. The researchers took 15 volunteers, sat each one in front of a regular-size 15-inch monitor and had them complete a variety of tasks designed to challenge their powers of concentration - like a Web search, some cutting and pasting and memorizing a seven-digit phone number. Then the volunteers repeated these same tasks, this time using a computer with a massive 42-inch screen, as big as a plasma TV.

The results? On the bigger screen, people completed the tasks at least 10 percent more quickly - and some as much as 44 percent more quickly. They were also more likely to remember the seven-digit number, which showed that the multitasking was clearly less taxing on their brains. Some of the volunteers were so enthralled with the huge screen that they begged to take it home. In two decades of research, Czerwinski had never seen a single tweak to a computer system so significantly improve a user's productivity. The clearer your screen, she found, the calmer your mind. So her group began devising tools that maximized screen space by grouping documents and programs together - making it possible to easily spy them out of the corner of your eye, ensuring that you would never forget them in the fog of your interruptions. Another experiment created a tiny round window that floats on one side of the screen; moving dots represent information you need to monitor, like the size of your in-box or an approaching meeting. It looks precisely like the radar screen in a military cockpit.

In late 2003, the technology writer Danny O'Brien decided he was fed up with not getting enough done at work. So he sat down and made a list of 70 of the most "sickeningly overprolific" people he knew, most of whom were software engineers of one kind or another. O'Brien wrote a questionnaire asking them to explain how, precisely, they managed such awesome output. Over the next few weeks they e-mailed their replies, and one night O'Brien sat down at his dining-room table to look for clues. He was hoping that the self-described geeks all shared some common tricks.

He was correct. But their suggestions were surprisingly low-tech. None of them used complex technology to manage their to-do lists: no Palm Pilots, no day-planner software. Instead, they all preferred to find one extremely simple application and shove their entire lives into it. Some of O'Brien's correspondents said they opened up a single document in a word-processing program and used it as an extra brain, dumping in everything they needed to remember - addresses, to-do lists, birthdays - and then just searched through that file when they needed a piece of information. Others used e-mail - mailing themselves a reminder of every task, reasoning that their in-boxes were the one thing they were certain to look at all day long.

In essence, the geeks were approaching their frazzled high-tech lives as engineering problems - and they were not waiting for solutions to emerge from on high, from Microsoft or computer firms. Instead they ginned up a multitude of small-bore fixes to reduce the complexities of life, one at a time, in a rather Martha Stewart-esque fashion.

Many of O'Brien's correspondents, it turned out, were also devotees of "Getting Things Done," a system developed by David Allen, a personal-productivity guru who consults with Fortune 500 corporations and whose seminars fill Silicon Valley auditoriums with anxious worker bees. At the core of Allen's system is the very concept of memory that Mark and Czerwinski hit upon: unless the task you're doing is visible right in front of you, you will half-forget about it when you get distracted, and it will nag at you from your subconscious. Thus, as soon as you are interrupted, Allen says, you need either to quickly deal with the interruption or - if it's going to take longer than two minutes - to faithfully add the new task to your constantly updated to-do list. Once the interruption is over, you immediately check your to-do list and go back to whatever is at the top.

"David Allen essentially offers a program that you can run like software in your head and follow automatically," O'Brien explains. "If this happens, then do this. You behave like a robot, which of course really appeals to geeks."

O'Brien summed up his research in a speech called "Life Hacks," which he delivered in February 2004 at the O'Reilly Emerging Technology Conference. Five hundred conference-goers tried to cram into his session, desperate for tips on managing info chaos. When O'Brien repeated the talk the next year, it was mobbed again. By the summer of 2005, the "life hacks" meme had turned into a full-fledged grass-roots movement. Dozens of "life hacking" Web sites now exist, where followers of the movement trade suggestions on how to reduce chaos. The ideas are often quite clever: O'Brien wrote for himself a program that, whenever he's surfing the Web, pops up a message every 10 minutes demanding to know whether he's procrastinating. It turns out that a certain amount of life-hacking is simply cultivating a monklike ability to say no.

"In fairness, I think we bring some of this on ourselves," says Merlin Mann, the founder of the popular life-hacking site "We'd rather die than be bored for a few minutes, so we just surround ourselves with distractions. We've got 20,000 digital photos instead of 10 we treasure. We have more TV Tivo'd than we'll ever see." In the last year, Mann has embarked on a 12-step-like triage: he canceled his Netflix account, trimmed his instant-messaging "buddy list" so only close friends can contact him and set his e-mail program to bother him only once an hour. ("Unless you're working in a Korean missile silo, you don't need to check e-mail every two minutes," he argues.)

Mann's most famous hack emerged when he decided to ditch his Palm Pilot and embrace a much simpler organizing style. He bought a deck of 3-by-5-inch index cards, clipped them together with a binder clip and dubbed it "The Hipster P.D.A." - an ultra-low-fi organizer, running on the oldest memory technology around: paper.

In the 1920's, the Russian scientist Bluma Zeigarnik performed an experiment that illustrated an intriguing aspect of interruptions. She had several test subjects work on jigsaw puzzles, then interrupted them at various points. She found that the ones least likely to complete the task were those who had been disrupted at the beginning. Because they hadn't had time to become mentally invested in the task, they had trouble recovering from the distraction. In contrast, those who were interrupted toward the end of the task were more likely to stay on track.

Gloria Mark compares this to the way that people work when they are "co-located" - sitting next to each other in cubicles - versus how they work when they are "distributed," each working from different locations and interacting online. She discovered that people in open-cubicle offices suffer more interruptions than those who work remotely. But they have better interruptions, because their co-workers have a social sense of what they are doing. When you work next to other people, they can sense whether you're deeply immersed, panicking or relatively free and ready to talk - and they interrupt you accordingly.

So why don't computers work this way? Instead of pinging us with e-mail and instant messages the second they arrive, our machines could store them up - to be delivered only at an optimum moment, when our brains are mostly relaxed.

One afternoon I drove across the Microsoft campus to visit a man who is trying to achieve precisely that: a computer that can read your mind. His name is Eric Horvitz, and he is one of Czerwinski's closest colleagues in the lab. For the last eight years, he has been building networks equipped with artificial intelligence (A.I.) that carefully observes a computer user's behavior and then tries to predict that sweet spot - the moment when the user will be mentally free and ready to be interrupted.

Horvitz booted the system up to show me how it works. He pointed to a series of bubbles on his screen, each representing one way the machine observes Horvitz's behavior. For example, it measures how long he's been typing or reading e-mail messages; it notices how long he spends in one program before shifting to another. Even more creepily, Horvitz told me, the A.I. program will - a little like HAL from "2001: A Space Odyssey" - eavesdrop on him with a microphone and spy on him using a Webcam, to try and determine how busy he is, and whether he has company in his office. Sure enough, at one point I peeked into the corner of Horvitz's computer screen and there was a little red indicator glowing.

"It's listening to us," Horvitz said with a grin. "The microphone's on."

It is no simple matter for a computer to recognize a user's "busy state," as it turns out, because everyone is busy in his own way. One programmer who works for Horvitz is busiest when he's silent and typing for extended periods, since that means he's furiously coding. But for a manager or executive, sitting quietly might actually be an indication of time being wasted; managers are more likely to be busy when they are talking or if PowerPoint is running.

In the early days of training Horvitz's A.I., you must clarify when you're most and least interruptible, so the machine can begin to pick up your personal patterns. But after a few days, the fun begins - because the machine takes over and, using what you've taught it, tries to predict your future behavior. Horvitz clicked an onscreen icon for "Paul," an employee working on a laptop in a meeting room down the hall. A little chart popped up. Paul, the A.I. program reported, was currently in between tasks - but it predicted that he would begin checking his e-mail within five minutes. Thus, Horvitz explained, right now would be a great time to e-mail him; you'd be likely to get a quick reply. If you wanted to pay him a visit, the program also predicted that - based on his previous patterns - Paul would be back in his office in 30 minutes.

With these sorts of artificial smarts, computer designers could re-engineer our e-mail programs, our messaging and even our phones so that each tool would work like a personal butler - tiptoeing around us when things are hectic and barging in only when our crises have passed. Horvitz's early prototypes offer an impressive glimpse of what's possible. An e-mail program he produced seven years ago, code-named Priorities, analyzes the content of your incoming e-mail messages and ranks them based on the urgency of the message and your relationship with the sender, then weighs that against how busy you are. Superurgent mail is delivered right away; everything else waits in a queue until you're no longer busy. When Czerwinski first tried the program, it gave her as much as three hours of solid work time before nagging her with a message. The software also determined, to the surprise of at least one Microsoft employee, that e-mail missives from Bill Gates were not necessarily urgent, since Gates tends to write long, discursive notes for employees to meditate on.

This raises a possibility both amusing and disturbing: perhaps if we gave artificial brains more control over our schedules, interruptions would actually decline - because A.I. doesn't panic. We humans are Pavlovian; even though we know we're just pumping ourselves full of stress, we can't help frantically checking our e-mail the instant the bell goes ding. But a machine can resist that temptation, because it thinks in statistics. It knows that only an extremely rare message is so important that we must read it right now.

So will Microsoft bring these calming technologies to our real-world computers? "Could Microsoft do it?" asks David Gelernter, a Yale professor and longtime critic of today's computers. "Yeah. But I don't know if they're motivated by the lust for simplicity that you'd need. They're more interested in piling more and more toys on you."

The near-term answer to the question will come when Vista, Microsoft's new operating system, is released in the fall of 2006. Though Czerwinski and Horvitz are reluctant to speculate on which of their innovations will be included in the new system, Horvitz said that the system will "likely" incorporate some way of detecting how busy you are. But he admitted that "a bunch of features may not be shipping with Vista." He says he believes that Microsoft will eventually tame the interruption-driven workplace, even if it takes a while. "I have viewed the task as a 'moon mission' that I believe that Microsoft can pull off," he says.

By a sizable margin, life hackers are devotees not of Microsoft but of Apple, the company's only real rival in the creation of operating systems - and a company that has often seemed to intuit the need for software that reduces the complexity of the desktop. When Apple launched its latest operating system, Tiger, earlier this year, it introduced a feature called Dashboard - a collection of glanceable programs, each of which performs one simple function, like displaying the weather. Tiger also includes a single-key tool that zooms all open windows into a bingo-card-like grid, uncovering any "lost" ones. A superpowered search application speeds up the laborious task of hunting down a missing file. Microsoft is now playing catch-up; Vista promises many of the same tweaks, although it will most likely add a few new ones as well, including, possibly, a 3-D mode for seeing all the windows you have open.

Apple's computers have long been designed specifically to soothe the confusions of the technologically ignorant. For years, that meant producing computer systems that seemed simpler than the ones Microsoft produced, but were less powerful. When computers moved relatively slowly and the Internet was little used, raw productivity - shoving the most data at the user - mattered most, and Microsoft triumphed in the marketplace. But for many users, simplicity now trumps power. Linda Stone, the software executive who has worked alongside the C.E.O.'s of both Microsoft and Apple, argues that we have shifted eras in computing. Now that multitasking is driving us crazy, we treasure technologies that protect us. We love Google not because it brings us the entire Web but because it filters it out, bringing us the one page we really need. In our new age of overload, the winner is the technology that can hold the world at bay.

Yet the truth is that even Apple might not be up to the task of building the ultimately serene computer. After all, even the geekiest life hackers find they need to trick out their Apples with duct-tape-like solutions; and even that sometimes isn't enough. Some experts argue that the basic design of the computer needs to change: so long as computers deliver information primarily through a monitor, they have an inherent bottleneck - forcing us to squeeze the ocean of our lives through a thin straw. David Rose, the Cambridge designer, suspects that computers need to break away from the screen, delivering information through glanceable sources in the world around us, the way wall clocks tell us the time in an instant. For computers to become truly less interruptive, they might have to cease looking like computers. Until then, those Post-it notes on our monitors are probably here to stay.

Clive Thompson is a contributing writer for the magazine.

Copyright 2005 The New York Times Company

Friday, October 14, 2005

From disorganization into organization: of language and biology.

Perspectives in Biology and Medicine 48.3 (2005) 317-327

"Meaning-Making" in Language and Biology

Yair Neuman

Ben-Gurion University of the Negev, Beer-Sheva, Israel.


The linguistic metaphor in biology adheres to a representational theory that seeks similarities between pre-given domains. The point of departure of this paper is the generative and nonrepresentational conception of metaphor. This paper argues that by adopting the nonrepresentational conception of metaphor, meaning-making may be the appropriate perspective for understanding biological systems. In both cases (the linguistic and the biological), boundary conditions between different levels of organization use micro-level disorganization to create macro-level organization.

Human Thinking Is Metaphorical, and metaphors are the sine qua non of any process of understanding (Lakoff and Johnson 1980). Thus, a scientific theory must critically examine its reservoir of metaphors and seek alternative metaphors for enlarging its scope and transcending its boundaries. As Tauber (1996) argues, "Theory must grope for its footing in common experience and language. By its very nature the metaphor evokes and suggests but cannot precisely detail the phenomenon in concern" (p. 18). Indeed, metaphors are creatively generated rather than mechanically applied to a pre-given world (Shanon 1992), and therefore they cannot "detail the phenomenon in concern"—an activity which is the role of the scientific model—but only guide the inquiry.

Human language has been used metaphorically to understand biological [End Page 317] processes, and vice versa (e.g., Atlan and Cohen 1998; Novak and Komarova 2001). This paper concerns the linguistic metaphor in biology. Can the linguistic metaphor productively guide investigation into biological systems? To address this difficult issue, we should be familiar with an important distinction between the representational and the nonrepresentational approaches to metaphor.

As Shanon argues, the discussion of metaphors in cognition and related disciplines commonly assumes that metaphor is a relationship established between two given entities whose attributes are defined prior to the establishment of their relationship. This representational theory of metaphor is evident in Gentner's (1983) seminal work on metaphor as a form of structural mapping between two domains. For example, the metaphor "An atom is like the solar system" is interpreted as a mapping of known, deep-structure similarities (similar relations) between one domain (the atom) and another (the solar system); electrons revolve around the nucleus just as the earth revolves around the sun. Although in some cases the use of metaphors may be interpreted by the representational theory, Shanon propounds the alternative that in most cases a metaphor has generative power that creates the similarities rather than simply assuming them. That is, the "metaphoric relationship is more basic than its constituents," and the metaphor "creates new features and senses" (Shanon 1992, p. 674).

The linguistic metaphor in biology has mainly employed the representational theory of metaphor to look for similarities between human language and biological systems as two pre-given domains. The benefit of moving along this line of inquiry is questionable. If one is familiar with the pre-given properties of two domains, finding similarities between them is of little use. Indeed, students of linguistics do not have to read Essential Cell Biology in order to understand human language, and students of medicine do not need to master Chomsky to understand cell biology. This critique of the representational theory of metaphor, while not new to those familiar with theories of metaphor, casts serious doubt on the possible contribution of the linguistic metaphor to biology.

If we adopt the nonrepresentational approach, our strategy should be different: first we should draw the metaphor and only then examine the similarities that emerge from its use. To illustrate this strategy, let us examine a difficulty that results from the representational approach to the linguistic metaphor.

The linguistic metaphor in biology has focused almost exclusively on similarities between the syntax of linguistic and biological systems for example, as evident in the structure of DNA. This is no surprise. Our knowledge of syntax has reached a high level of abstraction and formality that makes it easy to draw the analogy between the syntax of language and the "syntax" of DNA. However, the scope of linguistics is much broader than the study of syntax, and in order to have a full grasp of a linguistic activity one must also study, for example, the pragmatics of language (Yule 1998). Pragmatics is a field of linguistics that deals with language usage in context—in other words, the field of linguistics that deals with the generation of meaning-in-context (Levinson 1998). Although the generation [End Page 318] of meaning-in-context is crucial for understanding biological systems (Neuman 2004a), the linguistic metaphor in biology has, for the most part, ignored pragmatics. For example, Ji (1997), who propounds the idea of cell language, describes human language as consisting of lexicon, grammar, phonetics/phonology, and semantics, but he ignores pragmatics. This may be explained by the tremendous difficulty facing pragmatics even in linguistics (Yule 1998). This difficulty, however, holds great promise for biology and linguistics/semiotics alike. The analogy between human language and biological systems may teach both biology and linguistics an important lesson on a difficult subject: how meaning emerges in context. This paper intends to make the first moves toward this kind of inquiry.

Living Systems and Boundary Conditions

Biological systems are open systems that exist on several distinct, complementary, and irreducible levels of organization. These levels constitute the systemic closure of the living system through feedback loops. In other words, they are recursive-hierarchical systems (Bateson 1979; Harries-Jones 1999). This unique form of organization, which also characterizes the structure of text, contrasts sharply with the organization of information-processing devices. Information, as classically defined by Shannon, is a probabilistic measure. As Emmeche and Hoffmeyer (1991) argue, unpredictable events are an essential part of life, and thus it is impossible to assign distinct probabilities to new events: "The quantitative concept of information needs a closed possibility space. If the set of possibilities is open, one cannot ascribe precise probabilities to any single possibility and thus no information value" (p. 3). Their conclusion is that biological information must embrace the "semantic openness" that is evident, for example, in human communication, and that we should abandon the probabilistic conception of information. Indeed, the semantic openness of language allows the free interplay of ideas and concepts, just as a certain level of disorganization in living systems is necessary for the emergence of new forms. Without a basic level of disorganization, semantic openness cannot exist.

Following Bateson (1979), Hoffmeyer and Emmeche (1991) also propound the idea that living systems have two different codes: a digital binary code for memory (as in DNA) and a gestalt-type analog code for behavior. The syntactic approach to language has emphasized the digital aspect without paying attention to the analog one. However, if we want to understand living systems as meaning-making systems, then the analog mode is indispensable. Indeed, several scholars have argued that living systems are reactive rather than transformatory (information-processing) (Cohen 2000; Neuman 2003). Transformational systems are sequential, linear systems that transform information in a specific order to achieve a specific goal (Cohen 2000). In contrast, reactive systems are multilevel, nonlinear, ongoing systems that interact constantly with their internal and external [End Page 319] environment to create sense out of the environment in an integrative gestalt manner that cannot be reduced to a digital binary code. In other words, living systems are meaning-making machines rather than information-processing devices.

The recursive-hierarchical and semantically open structure of living systems is evident in protein conformation. Although a protein folds to assume an energetically favorable structure, we cannot understand its final conformation without taking into account several distinct and complementary levels of organization and the boundary conditions imposed by the higher levels, as well as the context or environment—its interactions with solvent and with ligands. To understand protein folding, we must take into account not only different levels of organization but also interaction-in-context. Metaphorically speaking, we must take into account the "pragmatics" of this process.

The idea that the living organism is composed of an irreducible structure was introduced by Michael Polanyi (1968). One of Polanyi's main arguments is that an organism is a system whose structure serves as "a boundary condition harnessing the physical-chemical processes by which its organs perform their functions" (p. 1308). In other words, "if the structure of living things is a set of boundary conditions, this structure is extraneous to the laws of physics and chemistry which the organism is harnessing" (Polanyi 1968, p. 1309). If each level imposes a boundary on the operation of a lower level, then the higher level forms the meaning of the lower level, as evident in the folding of proteins.

Polanyi illustrates his thesis by analogy with linguistics. According to Polanyi, the boundary conditions in living systems are analogous to the boundary conditions in linguistics. The meaning of a word is determined by the sentence in which it is located, and the meaning of a sentence is determined by the text in which it is located. This analogy should be qualified. First, although we cannot understand the words without understanding the sentence, neither can we understand a sentence without understanding its words. This "hermeneutic circularity" was recognized long ago, and it seems to characterize the operation of living systems (Neuman 2003). However, due to a misunderstanding of the recursion process and the recursive-hierarchical organization of living systems, this hermeneutic circularity has been considered, at least by some philosophers, something to be avoided rather than a constitutive principle of living systems. Second, biological systems may be metaphorically described as "texts." However, there is no text without a reader. There is no meaning without an interaction. Therefore, both recursive-hierarchical organization and interaction are crucial for describing biological systems in linguistic terms.

The fact that Polanyi and several celebrated biologists (e.g., Atlan and Cohen 1998; Cohen 2000; Jerne 1984) use the linguistic metaphor for understanding biological systems is not due to an intellectual whim. Meaning-making in natural language and the behavior of living systems do have something is common: both take advantage of disorganization on the micro level to create organization [End Page 320] on the macro level, through a recursive-hierarchical structure and interaction. Both operate on the boundary of organization and disorganization to create meaning-in-context.


Meaning-making can be defined as a process that yields the system's differentiated response to an indeterminate signal (Neuman 2004a). For example, being an antigen is not an attribute that is explicitly or directly expressed by a molecule (i.e., the signal). The meaning of being an antigen is the result of a complex deliberation process (i.e., a meaning-making process) that is finally evident in the specific immune response (Cohen 2000). In this sense, meaning-making is a process of computation in the classical etymological sense of assembling a whole from pieces.

The term computation is usually used in the technical, modern sense of a deductive process following a deterministic algorithmic program. Von Foerster and Poerksen (2002) have suggested restoring the original meaning of this word: computation is derived from the Latin computare, where com means "together" and putare means "to contemplate" or "to consider." Meaning-making involves bringing together different perspectives to achieve a specific response. For example, the decision as to whether a specific agent is an antigen or not involves a variety of immune agents (macrophages, T cells, B cells, cytokines) that contemplate (putare) together (com) to yield the final immune response. In other words, the signal (e.g., an antigen) is contextualized in a wider network of immune agents to achieve a specific response (Neuman 2004a). Meaning-making is thus a process of computation in the analog, holistic, and gestalt senses.

If there are no degrees of freedom in the system's response to a given signal, then by definition this system is not involved in meaning-making. The potentiality of a signal is an important economical principle underlying communication processes in living systems. This flexibility may be illustrated through natural language, in which the same sign can be used in different contexts to express different things. The specific term for this phenomenon is polysemy. The benefit of polysemy is clear: it "allows the use of the same word in different contexts and thus endows language with indispensable flexibility" (Shanon 1993, p. 45). This point is crucial for understanding both meaning-making and the organization of living systems. In both cases, there is maximum potentiality (of the sign or the molecule) at the micro level that endows the system with tremendous flexibility for making sense (linguistic or biological) on the macro level. [End Page 321]

Meaning-Making, Organization, and Disorganization

The term sense may have different meanings and connotations in biology and linguistics. In this paper, I use the term as being closely associated with organization. Thus a signal (in biology) or a sign (in linguistics) has meaning, or makes "sense," if it is embedded in a higher-order structure of components (i.e., context) that enables the system to produce a specific response.

The interesting thing about meaning-making is its quasi-paradoxical nature and the fact that it operates on the boundary of organization and disorganization. Let me explain this argument in semiotic terms. The generation of meaning requires that the system have maximum freedom on the micro level, but minimum freedom on the macro level. That is, disorganization in the sense of flexibility and dynamics is a necessary component of meaning-making. In natural language, for example, we can produce an infinite number of "meanings" (responses, senses) with a limited number of words. Saussure pointed out that the meaning of a sign is determined by its location in a broader network of signs. In other words, the meaning of a sign is determined by different organizations of the signs among which it is contextualized. Thus, a "virgin" sign lacks any sense and can be linked to myriad organizations of signs; therefore its degrees of freedom are potentially limitless. A sign is always a potential before it is mapped onto the macro level of the sentence (Neuman 2003).

The above argument may be formulated in terms of Peirce's (1955) three modes of being. The first mode, "firstness," concerns pure potentiality. In the meaning-making process, it is associated with the most basic level of organization and with the potential variability of the signal. As Peirce argues: "Freedom can only manifest itself in unlimited and uncontrolled variety and multiplicity; and thus the first becomes predominant in the ideas of measureless variety and multiplicity" (p. 79). While "firstness" concerns pure potentiality, "secondness" deals with actualization of the potential through relations established between various components of the system. In this sense, secondness is a constraint—a form of organization—imposed on the first mode. In meaning-making, the second mode of being is associated with a process of contextualization in which a given sign is woven into a network of other signs. In immunology, secondness as a type of relationship is evident when a specific signal is patterned (contextualized) into the web of immune agents and through this context obtains its meaning as an antigen (Cohen 2000).

The third mode ("thirdness") is that "which is what it is by virtue of imparting a quality to reactions in the future" (Peirce 1955, p. 91). In other words, it is the law, the habit that governs the behavior of the phenomenon, and our ability to predict its future behavior based on the law. It is the "conception of mediation, whereby a first and second are brought into relation" (Peirce 1978, p. 25). In meaning-making, the third mode of being is evident when the system's different [End Page 322] perspectives converge and are integrated to achieve a specific response in a given context, a process described by myself (2004a), following Bakhtin, as "transgradience." In immunological recognition as a process of meaning-making, this mode of being is evident when the immune agents "co-respond" to each other through a complex communication network to reach the final decision as to whether a molecule is an antigen (Cohen 2000).

Meaning and Interaction

The movement from disorder (firstness) to order (thirdness) is not random. This point can be illustrated through protein folding. One might imagine that protein molecules search through all possible conformations at random "until they are frozen at the lowest energy in the conformation of the native state" (Branden and Tooze 1999, p. 91). However, this "random walk" would require far more than the actual folding time. In this sense, it is ridiculous to expect order on the macro scale of a living system to simply pop up from firstness, just as it is ridiculous to expect a meaningful theory to emerge out of the uncontrolled delirium of a schizophrenic patient. In the case of proteins, Branden and Tooze (1999) have suggested that "the folding process must be directed in some way through a kinetic pathway of unstable intermediates to escape sampling a large number of irrelevant conformations" (p. 91). As Peirce argues, one cannot explain meaning by reducing it to lower levels of analysis; meaning is evident only once there is a triadic (or higher-order) relation between components—in other words, only through the mediating force of thirdness. In this sense, intermediates are needed to produce sense from senseless microelements.

The protein-folding process is a riddle. It is known, however, that the decrease in free energy is not linear. During the folding process, the protein proceeds from a high-energy, unfolded state (high potentiality) to a low-energy, native state through "metastable intermediate states with local low energy minima separated by unstable transition states of higher energy" (Branden and Tooze 1999, p. 93). To understand this process in semiotic terms, think about a word. It is transformed from a state of high potentiality to concrete actuality through the regulatory power of the higher linguistic levels of analysis, such as a sentence or text. However, every move from one linguistic level of analysis to a higher level of analysis gives the word new potential for a different response. The word love, for example, has the potential to mean different things in different contexts. The potential is actualized and the meaning of love is constrained when it is in a sentence. However, placing the sentence in a broader, extra-linguistic context may result in a totally different response than the one expected from the sentence alone. In the context of protein folding, this process suggests that between the micro and the macro levels of analysis there is a process of organization—one that has been described as the "middle way" or the "logic of in-between" [End Page 323] (Laughlin et al. 2000; Neuman 2004b). This process seeks to overcome high-energy barriers to folding (constraints). Producing order from chaos is energy-consuming, but without the system's basic tendency to reduce its free energy or to revert to a more basic mode of being, no work can be done and no meaning can be created. In this sense, the protein's natural entropic path to disorganization is subject to meta-regulatory processes (boundary conditions) that channel it so as to increase order.

These meta-stable processes can be discussed in terms of interaction. The transfer of energy involves a weak coupling/interaction between at least two systems (e.g., ligand-receptor binding). This coupling is weak in the sense of the weak interactions discussed in research on synchronization. That is, one system/level of organization transfers energy to the other system/level of organization, but they remain two autonomous and separate systems/levels of organization. Weak interactions are a necessary condition for meaning-making, whether in biology (e.g., non-covalent forces) or in linguistics. Returning to Peirce, we can understand that the pure potentiality of the micro level is actualized by the repeated, habitual, or synchronized interaction of the third level. It is the third level that completes the triadic structure of meaning-making. Interaction is what mediates the emergence of meaning, whether in linguistics or in biology.

Although meaning-making assumes disorder, the specificity of response demands order on the macro level. When a sign is located in a context, its degrees of freedom are significantly reduced. Thus, we need minimal potentiality on the macro level. For example, when using the sign shoot, we would like to have maximum freedom to use the same word once to express an order given to a soldier, and in a different context as a synonym for speak. In a given context, however, we want the sign to communicate only one of the meanings and not the other. To achieve this stability, the system has to be habituated through socialization, in the case of human learning of signs, or through evolutionary processes, in the case of a biological response. In both cases, interaction plays a crucial role. One must pay close attention to the fact that this "habituation" is not domestication. When we use language, we would like to be understood through the social habituation of language practices. On the other hand, we always preserve the opportunity to be incomprehensible and vague through jokes, paradoxes, inventions, and other forms of nonsense that are crucial for flexible life.

The need for macro stability is evident in protein conformation. A given sequence of amino acids forms a stable structure through the covalent forces that bind its molecules. However, this order is subject to non-covalent forces that interact to yield a specific conformation. Although a protein has an enormous number of potential conformations, it finally folds into one main conformation capable of responding (for example) to a given ligand. In other words, the final conformation of the protein is determined through interaction with another biological entity in a given context (Cohen 2000). This description should be qualified, too. Although meaning-making requires stability at the macro level, it [End Page 324] is not a total, rigid stability. The function of a protein requires structural flexibility and not rigid stability. The function of a word requires the same flexibility. You may want your word to achieve a specific response in a specific context, but you always want to reserve the ability to take your word back. Both in life and language, pragmatics assumes flexibility.


Previous uses of the linguistic metaphor in biology have been based on a representational theory. I propose investigating biological systems through another perspective, which emphasizes the pragmatic and contextual side of language. If we are prepared to do this, then our next step is to discuss several aspects of biological meaning-making along the lines we drew previously. First, we should emphasize the idea of biological organization rather than biological order. Whereas the term order usually pertains to information theory and the idea that a phenomenon can be represented (and quantified) through a unidimensional string of characters, organization emphasizes the multi-level structure of biological systems and the interconnections among the components of the system. As Denbigh (1989) argues, wallpaper with a repeating pattern may be highly ordered but poorly organized. In contrast, a painting by Cezanne has a low level of order but a high (but not quantifiable) level of organization. Language usage is organized rather than ordered. Any attempt to reduce meaning to order is doomed to failure. Following this line of reasoning, we may interpret the pathology of a biological system, as in the case of autoimmune diseases, as a problem of disorganization and meaning-making—a problem with the system's ability to make sense out of signals by patterning them into a broader network of meaning. In this case, something happens in between the levels of organization; the boundary conditions do not function in such a way as to avoid the system's natural "entropic" reversion to firstness.

As Harries-Jones (n.d.) argues, following Bateson, the death of a living system is more likely to be related to its loss of flexibility/resilience and to the devastation of its capacity to self-organize than to an outright loss of energy. The death of an organism might be the result of a vicious pathogen, but blaming the pathogen is of no help. This phenomenon should be investigated primarily through the failure of the immune system to make sense out of signals. For example, cytokines have been considered crucial to immune recognition. Under certain conditions, however, knocking out genes responsible for the production of cytokines does not destroy the immune system (Cohen 2000). This is not a surprising finding if one realizes that biological systems in general are characterized by overlapping and redundant feedback mechanisms. In this specific case, the immune system had organized itself to function properly. In other words, the immune system shows resilience, in the sense that it renews itself and can therefore flexibly use other opportunities for meaning-making and immune recognition (Neuman 2004a). [End Page 325]

A mature understanding of a living system is possible only when the descriptions of the different levels are synthesized into a working whole. This is evident in immunology, where we have acquired knowledge about micro elements and their interaction in the immune system but without an encompassing synthesis (Cohen 2000). For example, cytokines play an important role in communication among immune agents, and there is a flood of information about cytokines. However, "practically nothing is known about the behavior of the [cytokine] network as a whole" (Callard, George, and Stark 1999). This problem should not be underestimated. As Paton (2000, p. 63) argues, "From a biological system's point-of-view there is a lack of tools of thought for dealing with integrative issues." Unless several distinct but complementary levels of organization are integrated and it is shown how they influence each other, the behavior of the immune system is to a large extent incomprehensible. This conclusion is, in fact, an invitation for researchers to investigate the recursive-hierarchical structure of living systems and the unique way these systems make sense out of their environment.

The author wishes to thank Irun Cohen and Peter Harries-Jones for ongoing discussions and their constant support, Jesper Hoffmeyer for his constructive comments on an earlier draft, and the anonymous reviewers for their helpful comments.

Atlan, H., and I. Cohen. 1998. Immune information, self-organization and meaning. Int Immunol 10:711–17.

Bateson, G. 1979. Mind and nature. New York: Dutton.

Branden, C., and J. Tooze. 1999. Introduction to protein structure. New York: Garland.

Callard, R., A. J. George, and J. Stark. 1999. Cytokines, chaos, and complexity. Immunity 11:507–13.

Cohen, I. R. 2000. Tending Adam's garden. New York: Academic Press.

Denbigh, K. G. 1989. Note on entropy, disorder, and disorganization. Br J Phil Sci 40:323– 33.

Emmeche, C., and J. Hoffmeyer. 1991. From language to nature. Semiotica 84:1–42.

Gentner, D. 1983. Structure-mapping: A theoretical framework for analogy. Cognit Sci 7: 155–70.

Harries-Jones, P. 1999. A recursive-vision: Ecological understanding and Gregory Bateson. Toronto: Univ. of Toronto Press.

Harries-Jones, P. n.d. Re-generating environment: The logic of moving boundaries. In Re-embedding the economy in nature and society: A Polanyian approach to the environment, ed. J. G. Vaillancourt and P. Hamel. Montreal: Black Rose Books. Forthcoming.

Hoffmeyer, J., and C. Emmeche. 1991. Code-duality and the semiotics of nature. In Semiotic modeling, ed. M. Anderson and F. Merrell, 117–66. New York: Mouton de Gruyter.

Jerne, N. K. 1984. The generative grammar of the immune system. Science 229:1057–59.

Ji, S. 1997. Isomorphism between cell and human languages: Molecular biological, bioinformatics and linguistic implications. BioSystems 44:17–39.

Lakoff, G., and M. Johnson. 1980. Metaphors we live by. Chicago: Univ. of Chicago Press.

Laughlin, R. B., et al. 2000. The middle way. Proc Natl Acad Sci USA 97:32–37.

Levinson, S. 1996. Pragmatics. Cambridge: Cambridge Univ. Press.

Neuman, Y. 2003. How deep is the surface: A theoretical framework for understanding meaning-making in living systems. Foundations Sci 8:317–64. [End Page 326]

Neuman, Y. 2004a. Meaning-making in the immune system. Perspect Biol Med 47:317–27.

Neuman, Y. 2004b. Mirrors mirrored: Is that all there is? SEED J 4:58–69.

Novak, M. A., and N. L. Komarova. 2001. Toward an evolutionary theory of language. Trends Cognit Sci 5:288–95.

Paton, R. 2002. Process, structure and context in relation to integrative biology. BioSys-tems 64:63–72.

Peirce, C. S. 1955. The principles of phenomenology. In Philosophical writings of Peirce, ed. J. Buchler, 74–98. New York: Dover.

Peirce, C. S. 1978. First, second, and third. In Collected papers of Charles Sanders Peirce, vol. 6, ed. C. Hartshome and P. Weiss, 25–28. Cambridge: Harvard Univ. Press.

Polanyi, M. 1968. Life's irreducible structure. Science 160:1308–12.

Shanon, B. 1992. Metaphor: From fixedness and selection to differentiation and creation. Poetics Today 13:660–85.

Shanon, B. 1993. The representational and the presentational: An essay on cognition and the study of mind. London: Harvester-Wheatsheaf.

Tauber. A. I. 1996. The immune self: Theory or metaphor? Cambridge: Cambridge Univ. Press.

von Foerster, H., and B. Poerksen. 2002. Understanding systems: Conversations on epistemology and ethics. New York: Kluwer Academic/Plenum Publishers.

Yule, G. 1998. Pragmatics. Oxford: Oxford Univ. Press.

It all comes down to one question: what matters?

Perspectives in Biology and Medicine 47.4 (2004) 597-607

The Fundamentals of Clinical Discovery

Jonathan Rees

Grant Chair of Dermatology, University of Edinburgh, Room 4.018, First Floor, The Lauriston Building, Lauriston Place, Edinburgh, EH3 9HA, United Kingdom.



There is a widespread view that clinical research is failing to advance appropriately, particularly in comparison with other aspects of biomedical science. I argue that this is due in part to an inadequate understanding of how medical advance occurs. The common usage of such terms as basic or fundamental, or the uncritical use of the term model is unhelpful—unhelpful, in that such terms tend to presuppose a certain model of clinical advance that is unusual, and furthermore, because they tend to exaggerate the importance of research in subjects such as biochemistry and genetics at the expense of other areas. I suggest that much medical research is best viewed as a form of engineering rather than science, and that the knowledge base and research funding for the amelioration of disease needs to be much more broadly based than at present.

A steady stream of articles over the last 20 years has highlighted problems facing clinical research (Dieppe and Bachmann 2000; Goldstein and Brown 1997; Lenfant 2003; Rosenberg 1999; Snyderman 2004; Weatherall 1991). Most argue that undertaking clinical science is becoming more difficult, that funding is relatively harder to obtain, and that the career structures available for clinical researchers are inadequate. There is usually an implicit or explicit comparison with other forms of research which, for the moment, I will label as "basic" medical research. All these forms of research have as their principal justification the improvement of human health. Improvement in human health is [End Page 597] for instance seen as the chief justification for funding of the Human Genome Project and for the enormous increase in biomedical funding over the last 30 to 40 years (Greenberg 2001). How well these areas of rational enquiry map onto those needed to solve disease is a theme I will return to later in this essay.

The concern about the role of clinical science and clinical scientists in medical discovery has not been lost on some funders of medical research. A number of initiatives on both sides of the Atlantic have been undertaken recently to try and expand the role of clinical scientists in the medical research enterprise (Academy of Medical Sciences 2000). In the United Kingdom for instance, the main government funder of medical research, the Medical Research Council (MRC), has for the last several years sent letters to clinicians reminding them that the MRC still funds clinical science, although the examples they quoted as clinical science hardly reassure that the Council has much expertise in this area (MRC 2003). Also, in the United Kingdom, national academies, such as the recently formed Academy of Medical Sciences, have produced reports on how to encourage and develop the funding streams necessary for successful clinical science (Academy of Medical Sciences 2000). These efforts, however worthwhile, are in my view unlikely to shape events significantly (Rees 2002b). As I will argue below, the chief issue surrounding clinical science lies in our failure to grasp the nature of much clinically important discovery and in the promulgation of a worldview of medical research and advance that is dysfunctional and increasingly not fit for purpose (Rees 2002a, 2002b). Clinical research is in trouble, not primarily because of any lack of funds, but rather because lack of funds is merely a symptom of an incomplete understanding of clinical advance. In order to improve clinical science, we need to ask how we can foster advance and remove incentive structures that frustrate many aspects of the discovery process necessary for improved health care.

In what follows I will try to sketch some of the issues that I believe are important. I will briefly refer to some examples from an earlier article in Science (on the role of the study of complex genetics in medical research; Rees 2002a), but I will also put some of the arguments in a broader epistemological perspective. The nature of most science has changed in the last half century. The scientific enterprise has expanded enormously, becoming more expensive and increasingly specialized and complex. Central control, in terms of peer review and funding mechanisms, has also increased. Science and scientists have also become more partisan, with the accompanying results of a more widespread exaggeration of the importance of discoveries and a more short-term duration of the indicators of success. With its increase in size, the research community has seen a fragmentation of activity and the need to invent proxy measures of success and advance, so that the enterprise can be managed and funded and those carrying it out deemed responsible for public funds. At the same time, those carrying out most medical research have become ever more distant from the practice and delivery of health care. As much medical science has become expropriated from the clinical context, those undertaking it have frequently relied on secondhand and thirdhand descriptions of how health care is delivered and what represents genuine clinical advance. Finally, health care has become a major service sector in most Western economies, with both private companies and central governments becoming major players in providing technologies and the personnel involved in health care, parties that have obvious interests in the landscape of health delivery.

The Changing Nature and Practice of Science

Many of us have a romanticized (and beautiful) vision of scientific advance, a vision that might have been an accurate description of "revolutionary" science carried out by a small elite at a limited number of institutions half a century ago. My own favorite is in the celebrated account by Jim Watson (1999), in The Double Helix, of how he and Francis Crick reported (almost sadly it seems) getting only the occasional memo. Or think of Max Perutz's (1989) comments that he only had to start writing grant applications when he retired. How different that is from the everyday experiences of the present medical researcher. Grant applications are voluminous and increasingly—as Sydney Brenner (1996) points out—resemble documents that merely resemble flow diagrams of who reports to whom, with little room for science; ethics forms are subject to idiosyncratic criticism; and, in the United Kingdom at least, when you submit grant applications, it is unlikely that anybody around the grants committee table either knows you, has read your work at first hand, attends the same scientific meetings, or knows the views of your contemporaries in other countries. The community of peers has been replaced by a bureaucracy of proxy measures, such as grant income and impact factors.

Science has become more industrialized, with a rapid increase in the number of scientists and funding available to many individuals (Greenberg 2001; Ziman 1994, 2000). The size of many research groups has increased. In a wonderful obituary in Nature of Pat Wall (of the Melzack-Wall gate-theory fame), Clifford Woolf (2001) describes how modern lab heads are "really like chief executive officers of large multinational corporations, more involved in managing and delegating than in experimenting or thinking. Patrick David Wall, who died on 8 August aged 76, was the antithesis of this kind of scientist." There are very few persons who, like Fred Sanger, work literally with their own pair of hands. The increase in scale and cost of biological science has other implications. John Ziman, a former physicist, has chronicled the changing sociology of science (Ziman 1994, 1999, 2000). He points out that more and more science in universities has become "instrumental," undertaken as the production of knowledge with clearly foreseen or potential uses (Ziman 2002, 2003). Science is much less disinterested than it was: it has increasingly taken on many of the properties of commercial research and development activity—that is, it has become proprietary, more prosaic, pragmatic, and partisan. And I don't just mean that research funding often comes from industrial sources, but that, rather than comprising curiosity-driven research, the research agenda is managed and directed at many levels. Some diseases are determined to be more important that than others diseases, more worthy of funds. Patient lobby groups influence funding and prestige, and pharmaceutical companies are aware that their markets are likely to be bigger in some disease areas than others, or that major markets may lie outside the boundaries of what has traditionally been considered disease (e.g., cosmetic surgery). These factors are in part external to science but can be courted by scientists and groups of scientists with particular backgrounds and interests.

The clearest example of this trend has been provided by genetics. The emergence of the new genetic technologies, the ability to undertake genetics on man rather than just model organisms, rightly revolutionized much biomedical work. However, what has followed, as I have argued elsewhere (Rees 2002a), has been a genocentric view of medicine that has sought to concentrate funding and interest for a particular group of diseases and persons. Thus, after the hopelessly naïve view that identifying genes for Mendelian disorders would lead to therapy over a short period of time (rather than developing useful tools that allow the study of biology), genetics funding has followed the mantra of the need to understand complex diseases and gene therapy. But any sense of proportion has been rapidly lost. Insofar as talking about genetic and environmental causes makes sense—for example, for most cancers and most inflammatory diseases—only a relatively small portion of the variation seen in human populations is accounted for by genetic factors; thus, the impact of genetics on, for instance, the prediction of disease status for the majority of common diseases is going to be at best marginal, and far less important than other factors relating to heath care delivery. (Hemminki et al. 2001; Lichtenstein et al. 2000). By contrast, changes in incidence point clearly to the overriding importance of environmental factors for most common diseases. Leading journals have published review papers pointing out these views, yet the message appears to fall on stony ground (Holtzman and Marteau 2000; Weiss and Terwilliger 2000). A large number of mapping studies and association studies on common complex diseases are still published, and large population studies on common inflammatory conditions are still planned with the aim of identifying important genetic health determinants of common diseases. Yet we remain ignorant of the natural history of some of the most common inflammatory diseases of man, such as atopic dermatitis or psoriasis.

There is, understandably, a rush to use new technologies. Unfortunately, however, the attraction seems to overcome a sense of prudence about the likely rewards. If the pool of biomedical investigators trained in genetics is expanded, this expanded pool will continue to find problems to occupy them. They are unlikely to retrain to use other approaches to solve clinical problems. If you have spent time mapping the rare genes accounting for some rare cases of obesity, you are more likely to want to study the complex genetics of obesity than you are to start to study economics and look at the use of pricing to influence supply and demand of high calorie meals. When lobby groups launch a "Decade of the Brain" and train an increasing army of post-docs, are most of these scientists going to retrain in something else when the decade is over, or will the idea prevail that funding should be maintained or—better still—continually increased? A related problem is the way in which generic approaches appear to get adopted across whole swaths of medical research rather than in areas where they may be most useful. For instance, modern molecular technologies seem to hold out much hope for the diagnosis, management, and prevention of infectious disease. But the rises in obesity and type 2 diabetes are largely accounted for by changes in behavior: that we know, and the contribution of complex genetics would seem marginal to any sense of clinical reality.

Furthermore, there is often a general bias towards what sort of solution is to be sought. Here the increasing role of pharmaceutical industry in research goal setting and funding is important. Just as geneticists tend to do genetics, pharmaceutical companies sell drugs. They favor drug solutions for problems, as intellectual property is hard to obtain on other approaches to disease. And this bias affects not just directly funded work in universities but the tone of much medical research, where the promise of what may be patentable influences research direction. An example of this that I have used elsewhere is the case in which a genetic predictor of melanoma was favored over such easily observable markers of risk as skin color or freckles (Rees 2000).

Biology versus Medicine, and Biomedicine versus Health

The distinction between the understanding of the biology of disease and the knowledge required to improve the health of the population is an important one. The primary rationale for much biomedical research is that the reductionist enterprise that I associate with cell biology or biochemistry is the best way to improve health care. This view is, I believe, increasingly insular, partisan, and subject to challenge. Science, rightly, gains respect from its championing of a worldview that demands an external test in reality, the dialectic of theory against data. It would be ironic if it failed to apply the same standards to its own activities.

To follow my argument, we need examine some of the terms used to describe various forms of rational enquiry. Much biomedical research is fond of the description basic, or fundamental. Understanding the pathways in the cell is seen as basic to understanding (say diabetes). I take a different view. Medicine is not science—at least, it is not science in the way that most engineering is not science. Medicine itself I think of as a form of engineering: the design of systems of intervention or artifacts, based on underlying principles. And as in the case of engineering, the approach we take in medicine should be defined by the problem and context. Basic knowledge is that theory which provides the solution. If understanding the psychological factors that determine why people walk or ride by car a short distance influences body energy homeostasis, then that knowledge is basic.

Let me provide another topical example. In the United Kingdom, as in many other countries, there is a major shortage of cadaver kidneys for transplantation. One economist has suggested that providing a small tax incentive on everybody's tax return could save money by encouraging persons to register their organs, and by encouraging organ donation thus reduce the need for dialysis (Oswald 2001). Whether the idea will work, and the theory to test and experimentally resolve this issue, is basic knowledge: basic in the sense that we require it to solve the problem.

If we follow this argument through, we need to embrace a far wider range of academic disciplines as being relevant to medical care and to broaden the base of rational enquiry at the expense of the current staples. I have mentioned experimental economics, but critics will say that people already work in this area. In reality, the scale of activity is not commensurate with what is needed, nor is it commensurate with what is allocated to areas such as biochemistry.

One of the most important determinants of medical care is the doctor himself. The amount of work that examines the information processing and diagnostic skills of doctors is trivial compared with the funding made available for study of model organisms predicated on the assumption that this is the optimal way to improve health. Yet health informatics (including those relevant areas of biology and computing) is crucial to how medicine is practiced. And here I am making a plea not for pragmatic solutions to local issues, but for fundamental and theoretical work to underpin how those who diagnose are taught and kept competent. It is ironic that whereas disciplines such as informatics realize that study of both man and machine is necessary—that study of information handling by humans is relevant to how humans interact with machines and vice versa—medicine is still stuck in a Cartesian duality, believing that just delivering new drugs to the pharmaceutical salesman ensures improvement in health. Just as computing science embraces not just the physical world but the world of the human artifact, so should medicine (Simon 1969).

What Sort of Activity Is Clinical Science?

So far in this essay I have skirted around any definition of clinical science, although the skeptical reader will already have a fair idea of what I label as biochemistry or genetics. There are many definitions of clinical science. I particularly like the handshake test proposed by Goldstein and Brown (1997) for patient-oriented research, as it describes most of what I do. Do you need to shake hands with your subjects? If you do, then you are likely to be engaged in patient-oriented research. Nonetheless, whatever the tactical merits of using such a definition (and I think there are many), it appears too facile to accept the "conventional wisdom" and imagine that much clinical research is translational in nature, the mere testing of ideas developed in the laboratory. Clinical science is all too easily seen as journeyman work, unoriginal and merely applied. I think this view is mistaken. First, I would argue that the major direction of information flow is from patients to the laboratory. The wonderful unfolding of our knowledge about, say, structural proteins in skin has fed advance in cell biology far more than vice versa. Clinical science is, to put it slightly provocatively, more basic than cell biology. Mapping of a myriad of human disorders aids biology more than genetics has improved care. This is not to argue against the need for or the merits of much research activity, but rather to point out the inadequacy of the term translation. The idea of translation too often implies that the intellectual landscape has been defined in the "basic laboratory." Again, I disagree. The major issues facing medicine throughout much of the world relate to how we organize, deliver, and value the benefit health care provides.

Skeptics should look at the work of psychologist Daniel Kahneman, the 2003 Nobel Prize winner in economics, to see how our worldview, or what I might call our standard operating models for assessing disease, needs revising (Kahneman 2003; Kahneman and Tversky 1996; Redelmeier and Kahneman 1996; Redelmeier, Katz, and Kahneman 2003). Kahneman's work, we can safely say, is not mere humdrum recording what people say, but, as the Nobel committee realized, theory driven and empirically tested science that changes the research landscape irrevocably. These fields of knowledge are not mere "bolt-on" activities that you dream up when you do a clinical trial but areas that require basic intellectual ideas—blue sky thinking. We need much more empirically tested theoretical work around health care delivery and how medicine works, not merely pragmatic trials of one drug versus another in terms of whether "quality of life" improves.


A standard response to many of the issues I raise is, "But if we just knew in greater detail how the human genome worked (and while we are at it, the mouse, zebrafish, rat, worm, chimp genomes, too), and knew more cell biology, we could design rational therapies without all the waste and costs associated with conventional large pharmaceutical development (Glassman and Sun 2004). This is then followed by a request for large amounts of money for—depending on the time and audience—genomics, phenomics, systems biology, computational biology, or stem cell science. But clinical advance is not like this.

Elsewhere I have documented how, over the space of 20 years, clinical science has dramatically changed the management of the major common dermatoses, acne, psoriasis and forms of dermatitis, and skin cancer (Rees 2002a). Most advances have involved the use of technologies developed without a clear purpose and the linking of these technologies to clinical problems. Often significant advance has come from one of a handful of individuals, sometimes over a short period of time of a year or two. A complete understanding of disease was not necessary—nor is it in any logical sense ever attainable. Rather, advance requires investigational agents, usually drugs, assays that are close to the patient, and the ability to design proof of concept experiments (Rees 2002a). And here we must be aware that the sociology of clinical advance differs from that all too often projected as the way of doing successful science. Often the discoverers were isolated, without track records in the area of their discovery. What they do seem to have in common are powers of observation coupled with the ability to test ideas within a framework of clinical practice. It seems clear that this pattern of clinical discovery is the norm. David Healy (1996-2000), in a series of wonderful books of interviews, has documented how real therapeutic advance occurred in the golden age of clinical neuroscience. More recent examples, from sildenafil to botulinum, all suggest that this is the main route of clinical advance in medicine.

So does all this mean reductionism or biochemistry or genetics is not needed? Not at all, it is just not sufficient. It is far more dangerous not to be a reductionist than it is to be one—but one should accept reductionism's limitations. Neuroscience provides a good example in this regard. Studying the brain means working at different levels: molecules and populations, synapses and social groups. David Marr (1982), the late computational neuroscientist, said that trying to understand vision by studying only neurons was as futile as trying to understand bird flight by only studying feathers—it just can't be done. For medicine we need to be much more open to what sort of level of inquiry is needed in order to advance. The false dichotomy between basic (cell biology) and clinical science (translation) is unhelpful. Each level has to exist of itself, and the tools developed at one level may be useful to probe another, but often the attempt to map all down to the smallest scale will be unhelpful. Science is not a linear process from cell biology to population health but, by contrast, a series of activities and approaches, perhaps existing on the surface of a sphere as in non-Euclidian two-dimensional space, where the connotation of depth—implying deep or fundamental in a hierarchical sense—has no meaning.

"Basic" Revisited, and "Blue Skies" Research

Given the arguments I have made above, it is perhaps reasonable to ask whether I believe, or think it is sensible, to fund "blue-skies" research in the hope of promoting clinical care. If, as I have stated, basic is defined in terms of knowledge needed to solve a problem, is there still a space for disinterested work—curiosity-driven and perhaps playful intellectual activity? To answer this question, I need to return to some of the writings of John Ziman mentioned earlier, who has argued that academic science is increasingly partisan, pragmatic and less disinterested, and that it has taken on the properties of what was formerly described as industrial science (Ziman 2002).

There is, I believe, a confusion in the minds of many between basic (or fundamental) and disinterested research. It is too easy to say (for instance) that work on one of numerous model organisms is basic, and to confuse this with work that is blue skies (because the chance of any clinical relevance seems small). In reality, the work may have merits, even if the potential for clinical application is finite. The problem may be the mere use of the word model: one seems able to justify the study of all God's creatures this way. Of course, the understanding of any organism may have relevance to human health, but how does this approach compare with others? Kahneman's work on the memory of pain and discomfort, and on how humans view and make choices, arose not from an attempt to solve medical problems, but from efforts to study (among other things) decision making and judgment. I doubt that application or clinical relevance were in the minds of those trying to expose the weaknesses of the classical economic model of "rational man" (Lovallo and Kahneman 2003).

Another example would be the widespread use of computing and imaging in clinical medicine. Many of the technologies necessary were developed not with the goal of improving human health but for other reasons, not least a sense of intellectual curiosity. To me, these cases persuade that biochemistry and genetics are all too often conflated with curiosity-driven research at the expense of a broader intellectual horizon. Too frequently a "consensus" appears in which those arguing for funding seek to benefit from such funding, but often with little perspective on either clinical practice or how advance has happened previously.

A Sense for the Future

Whether the worldwide medical research landscape is getting more homogenous, with less tolerance of alternative approaches, I do not know. In the United Kingdom, as a result of a large mismatch between funding and those applying for funding, there is a tendency to concentrate funding in fewer centers. It seems that the harder financially starved universities bid for central funds, the more homogeneity and straitjacketing of research strategy. The consensus feeds on itself, with those benefiting the most in turn influencing policy of the limited number of funding streams. This may be, as supporters would argue, a way to maximize benefit from scarce funds; alternatively, as some of us believe, it may be a way to constrain advance and preserve the status quo. Freeman Dyson (1998) tells the story of how John Randall, a famous if not first-rate physicist, set in motion the seeds of what would become a major strength of British science in the mid-20th century, namely molecular biology. Whatever his own skills as a physicist, Randall had great insight into future strategies for understanding biology. But his success, and those of his students, relied on a highly decentralized system of science with individual autonomous institutions (even department heads) making their own decisions, and with a vision over decades: some succeeded, most failed. Perversely, while there is more money available today, the room for maneuver seems less.

There is no one clinical science, merely many ways in which the burden of disease can be ameliorated. We have failed to capitalize on many of those ways. A slight sense of dissatisfaction with the last quarter century of biomedical science is warranted. Medicine once again needs to redefine and broaden its intellectual heartlands (Rees 2002b).

The author would like to thank William Bains for helpful discussions.

Academy of Medical Sciences. 2000. The tenure-track clinician scientist: A new career pathway to promote recruitment into clinical academic medicine.

Brenner, S. 1996. Loose ends. Curr Biol 6(1):98.

Dieppe, P., and M. Bachmann. 2000. Clinical research: Is the filling missing from the medical research sandwich? J R Coll Physicians Lond 34(3):282-84.

Dyson, F. J. 1998. Imagined worlds. Cambridge: Harvard Univ. Press.

Glassman, R. H., and A. Y. Sun. 2004. Biotechnology: Identifying advances from the hype. Nature Reviews: Drug Discovery 3:177-83.

Goldstein, J. L., and M. S. Brown. 1997. The clinical investigator: Bewitched, bothered, and bewildered—but still beloved. J Clin Invest (12):2803-12.

Greenberg, D. S. 2001. Science, money, and politics: Political triumph and ethical erosion. Chicago: Univ. of Chicago Press.

Healy, D. 1996-2000. The psychopharmacologists: Interviews. 3 vols. London: Arnold.

Hemminki, K., et al. 2001. Estimation of genetic and environmental components in colorectal and lung cancer and melanoma. Gen Epidemiol 20(1):107-16.

Holtzman, N. A., and T. M. Marteau. 2000. Will genetics revolutionize medicine? N Engl J Med 343(2):141-44.

Kahneman, D. 2003. A perspective on judgment and choice: Mapping bounded rationality. Am Psychol 58(9):697-720.

Kahneman, D., and A. Tversky. 1996. On the reality of cognitive illusions. Psychol Rev 103 (3):582-91.

Lenfant, C. 2003. Clinical research to clinical practice: Lost in translation? N Engl J Med 349(9):868-74.

Lichtenstein, P., et al. 2000. Environmental and heritable factors in the causation of cancer: Analyses of cohorts of twins from Sweden, Denmark, and Finland [see comments]. N Engl J Med 343(2):78-85.

Lovallo, D., and D. Kahneman. 2003. Delusions of success. How optimism undermines executives' decisions. Harv Bus Rev 81(7):56-63, 117.

Marr, D. 1982. Vision: A computational investigation into the human representation and processing of visual information. New York: W. H. Freeman.

MRC. 2003. MRC: Clinical research news.

Oswald, A. J. 2001. Economics that matters: Using the tax system to solve the shortage of human organs. Kyklos 54:379-81.

Perutz, M. F. 1989. Is science necessary?: Essays on science and scientists. London: Barrie & Jenkins.

Redelmeier, D. A., and D. Kahneman. 1996. Patients' memories of painful medical treatments: Real-time and retrospective evaluations of two minimally invasive procedures. Pain 66(1):3-8. [End Page 606]

Redelmeier, D. A., J. Katz, and D. Kahneman. 2003. Memories of colonoscopy: A randomized trial. Pain 104(1-2):187-94.

Rees, J. 2000. Patents and intellectual property: A salvation for patient-orientated research? Lancet 356:849-50.

Rees, J. 2002a. Complex disease and the new clinical sciences. Science 296(5568):698- 700.

Rees, J. L. 2002b. Clinical Science: time for a new contract. Clin Med 2(5):423-25.

Rosenberg, L. E. 1999. The physician-scientist: An essential—and fragile—link in the medical research chain. J Clin Invest 103(12):1621-26.

Simon, H. A. 1969. The sciences of the artificial. Cambridge: MIT Press.

Snyderman, R. 2004. The clinical researcher: An "emerging" species. JAMA 291(7):882- 83.

Watson, J. D. 1999. The double helix. Middlesex: Penguin.

Weatherall, D. J. 1991. The physician scientist: An endangered but far from extinct species. BMJ 302:1002-5.

Weiss, K. M., and J. D. Terwilliger. 2000. How many diseases does it take to map a gene with SNPs? Nat Genet 26(2):151-57.

Woolf, C. J. 2001. Patrick D. Wall (1925-2001). Nature 413(6854):378.

Ziman, J. 1994. Prometheus bound: Science in a dynamic steady state. Cambridge: Cambridge Univ. Press.

Ziman, J. 1999. Rules of the game of doing science. Nature 400:721.

Ziman, J. 2000. Real science. Cambridge: Cambridge Univ. Press.

Ziman, J. 2002. The continuing need for disinterested research. Sci Eng Ethics 8(3):397- 99.

Ziman, J. 2003. Non-instrumental roles of science. Sci Eng Ethics 9(1):17-27.

from Booz Allen Hamilton: R&D Spending - as always, it matters how, not how much.

link to original article, including downloadable PDF reports at end.

No Relationship Between R&D Spending and Sales Growth, Earnings, or Shareholder Returns.

Analysis of the top 1,000 global innovation spenders finds accelerated investment in R&D—$384 billion in 2004.

NEW YORK, October 11, 2005 — There is no direct relationship between R&D spending and significant measures of corporate success such as growth, profitability, and shareholder return, according to a new global innovation study by Booz Allen Hamilton. However, the pace of corporate R&D spending continues to accelerate, as many executives continue to believe that enhanced innovation is required to fuel their future growth.

Booz Allen analyzed the world's top 1,000 corporate research and development spenders—the Booz Allen Global Innovation 1,000—to identify the linkages between spending on innovation and corporate performance, and to uncover insights on how organizations can get the greatest return on their innovation investment. Key findings of the study include:

Money doesn't buy results. While the study identified individual success stories, there is no discernable statistical relationship between R&D spending levels and nearly all measures of business success, including sales growth, gross profit, operating profit, enterprise profit, market capitalization, or total shareholder return.

R&D spending appears to yield better gross margins, the percentage of revenue left over after subtracting the direct costs incurred in making the products or services sold. This narrow departmental success, however, is not generally translated into overall corporate performance.
"There is no easy way to achieve sustained innovation success-you can't spend your way to prosperity," said Booz Allen Vice President Barry Jaruzelski. "Successful innovation demands careful coordination and orchestration both internally and externally. How you spend is far more important than how much you spend."

But innovation spending is still a growth business. The 2004 Global Innovation 1,000 spent $384 billion on R&D in 2004, representing 6.5% annual growth since 1999.

And the pace is accelerating-measured from 2002, the annual growth rate jumps to 11.0%.

Larger organizations have an advantage. Scale provides an edge in innovation; larger organizations are able to spend a smaller proportion of revenue on R&D than smaller organizations with no discernable impact on performance.

Spending more doesn't necessarily help, but spending too little will hurt. Companies in the bottom 10% of R&D spending as a percentage of sales under-perform competitors on gross margins, gross profit, operating profit, and total shareholder returns. However, companies in the top 10% showed no consistent performance differences compared to companies that spend less on R&D.

R&D spending by companies in developing nations is relatively small, but growing rapidly. While companies headquartered in North America, Europe, and Japan account for 96.8% of the Global Innovation 1,000's R&D spending, and are likely to remain dominant players for the foreseeable future, companies with headquarters in China, India, and the rest of the world are turning up the volume on R&D investment.

The annual growth rate for R&D spending from 1999 to 2004 in China and India was 21.1%, significantly higher than in North America (6.6%), Europe (6.2%), and Japan (4.8%). These lower growth rates are likely functions of the relative maturity of companies in these countries and the magnitude of their current spending.

However, the developed economies show a higher ratio of R&D spending to sales. Here China and India lag, spending only 1% of revenue on R&D, compared with 4.9% for firms in North America, 4% in Europe, and 3.8% in Japan. The differences among the three main spend regions are partially explained by differences in industry mix.

Industries can't agree on how much innovation spending is enough. Instead of clustering into any coherent pattern, R&D spending levels vary substantially, even within industries.

It's the process, not the pocketbook. Superior results seem to be a function of the quality of an organization's innovation process—the bets it makes and how it pursues them—rather than either the absolute or relative magnitude of its innovation spending. For example, Apple's 2004 R&D-to-Sales ratio of 5.9% trails the computer industry average of 7.6%, and its $489 million spend is a fraction of its larger competitors. But by rigorously focusing its development resources on a short list of projects with the greatest potential, the company created an innovation machine that eventually produced the iMac, iBook, iPod, and iTunes.

"The competitive value of a fast and effective innovation engine has never been greater," said Kevin Dehoff, Booz Allen Vice President, noting the trend toward shorter product life cycles and an ever-faster flow of new offerings. "Yet of all the core functions of most companies, innovation may be managed with the least rigor. The key is to identify the priority areas where process improvements will have the greatest impact."

Additional study findings include:

R&D spending is highly concentrated. While the top 1,000 corporate R&D spenders invested $384 billion in 2004, the second 1,000 spent only $26 billion-only an additional 6.8% beyond the top 1,000 spenders.

As a result, Booz Allen estimates that the Global 1,000 captures between 80-90% of total global corporate R&D spending, and approximately 60% of total global R&D, including spending by governments.

The top 10 global R&D spenders in 2004 are, in descending order: Microsoft, Pfizer, Ford, DaimlerChrysler, Toyota, General Motors, Siemens, Matsushita Electric, IBM, and Johnson & Johnson.

On average, the Global Innovation 1,000 spends 4.2% of its revenue on R&D. This average has been relatively stable over the last five years studied.

Patents don't always lead to profits. In a separate analysis, Booz Allen found no relationship between the number of patents issued to an organization and its performance.

R&D spending is heavily concentrated in the Technology, Health, and Automotive sectors. Computing & Electronics tops the list representing 25% of total spend by the Global 1,000; Health follows with 20%, and Auto with 18%.

Software & Internet, at 15% per annum, and Health at 12.4% have experienced the fastest pace of R&D growth over the past five years, while Telecom (2.2%) and Chemicals & Energy (1.4%) have grown the slowest.


Booz Allen Hamilton Global Innovation 1,000: Study Methodology

Booz Allen Hamilton identified the top 1,000 global public companies in Research and Development spending in 2004 for companies reporting their spend. The study then analyzed data for the past six years on a variety of financial metrics, including: revenue, gross profit, SG&A expenses, operating profit, net profit, capital expenditures, and historical R&D spending. The study also assessed shareholder value measures including total shareholder return (TSR) and market value growth for the same time period.

Each company was coded into one of 10 industry sectors or "other" and six country/regions based on Bloomberg's industry designations and reported headquarters locations, since R&D spending is only rarely broken out by subsidiary or region in corporate financial statements. For example, R&D spending for Chrysler conducted in the United States would be reported in Europe, given DaimlerChrysler's German headquarters.

To enable meaningful comparisons across industries on R&D spending levels, Booz Allen indexed the R&D spending level for each industry against the median R&D spending level for that industry. Similarly, to avoid skewing shareholder returns analysis by differences in performance across regional stock markets, shareholder returns data was adjusted to show the corporation's performance relative to that of a leading index of its regional market.