It is obvious why specialization improves productivity for manufactured goods (even if people seem to systematically underestimate just how extensive the gains are). If it takes 18 steps to manufacture a pin (to use Adam Smith's example), you can well imagine dividing up those steps among many workers to increase the number of pins produced. In a modern factory, workers are often further removed from the goods they are producing, but assembly lines demand exquisitely specialized machinery and workers well-trained in the operation of those machines. The nature of many manufacturing jobs has changed a great deal since the Industrial Revolution, but the same fundamental principles apply. Automation may have delegated much of the direct labor in factory to machines, but these machines nonetheless require skilled human beings to operate them.
In Book I, Chapter 1 of The Wealth of Nations, however, Smith notes that different sectors of the economy vary in the degree to which labor can be divided and in the ways in which the labor is divided. He notes especially that agriculture can afford limited specialization in tasks relative to other industries because the tasks (planting, harvesting, etc.) are performed seasonally rather than simultaneously. A pure specialist in planting would have little to do for much of the year, once the planting season is complete, and a pure specialist in harvesting would stand idle all year until the plants are fully grown. This basic pattern, evident in Smith's day, can still be observed today. I'm a country boy, after all, and I remember for many years watching the same farmer (and later his children) riding various machines around the field next door. My neighbor instead specialized in his crops, which were limited to corn, soybeans, and alfalfa (and never in the same year).
But our modern economy also features a dizzying array of "knowledge workers" – workers whose job is focused on producing, interpreting, and implementing our understanding of the world. This includes academic and industrial scientists, as well as many high-level engineers and designers and many others whose job description is hard to pin down in a word. Even back in 1776, Smith acknowledged the division of labor among knowledge workers:
In the progress of society, philosophy or speculation [NB: Smith here means knowledge pursuits in general, especially science and engineering] becomes, like every other employment, the principal or sole trade and occupation of a particular class of citizens. Like every other employment too, it is subdivided into a great number of different branches, each of which affords occupation to a peculiar tribe or class of philosophers; and this subdivision of employment in philosophy, as well as in every other business, improves dexterity, and saves time. Each individual becomes more expert in his won peculiar branch, more work is done upon the whole, and the quantity of science is considerably increased by it. [WN I.1.9].
Whereas it is not obvious that individual workers in manufacturing or agriculture are more specialized relative to Smith's day, any scientist can see plain as day that scientific specialization has exploded since those words were written. Smith himself "specialized" not only in economics, but also in moral philosophy, law, and even astronomy. With the possible exception of moral philosophy, all of these fields are today subdivided into a dozen or more sub-disciplines, and each of those sub-disciplines in turn into distinct research communities.
In psychology, this was brought to my attention forcefully when I read an article from the field of Science Education that alluded to an entirely separate research literature in Public Understanding of Science. Evidently the Science Education people barely talk to, attend conferences with, or even read the publications of people in Public Understanding of Science, and vice versa. Move over, Montagues and Capulets! For better and for worse, the division of labor has reached perhaps its fullest development in science.
Do scientists divide labor more like factory workers, each working on a specialized task (straightening the wire of a pin, executing experiments)? Or do they divide labor more like farmers, performing a diverse array of tasks organized around a specialized output (a crop of corn or a study of a particular topic)?
Scientists are more like farmers. For example, in my own field of experimental psychology, the same researcher often determines the research question (roughly, a prioritization task), reads the relevant research literature (a knowledge integration task), generates hypotheses (often a highly creative task), determines the experimental methodology (a logical reasoning task), supervises the execution of the experiment (a managerial task), analyzes the data (a software and statistics task), prepares the manuscript for publication (a written communication task), presents the work at conferences (an oral communication task), and begs for grant money for the next project (a persuasion task). An effective researcher in psychology needs to be competent at all of these tasks, and excellent at some of them. It would be utterly overwhelming, except that most psychologists study only a very narrow slice of questions (my own dilettantary is not representative).
Admittedly, psychology may be at the extreme, in part because psychology labs tend to be small relative to other, more mature disciplines that have more thoroughly proven their worth to society. For example, one of my friends is a molecular biologist who works in a lab that employs a dedicated staff member whose job is revising research articles written by graduate students and post-docs before they reach the desk of the lab's illustrious (and very busy) lead researcher. Likewise, in clinical trials for pharmaceuticals, it is common to have a dedicated statistician on the research team. In that most mature science of all, physics, there has long been a division between theorists and experimentalists.
Nonetheless, even though these more mature fields are more "vertically" integrated (in terms of personnel functions and tasks), this is nothing compared to their degree of "horizontal" integration (in terms of research topics or outputs). Looking through the faculty research interests at a top biology department, I find at random (not cherry-picking) a lab that studies cellular aging in yeast, one that studies the structure of the nuclear envelope during cell division, another focused on the functions of nucleic acids, and one investigating the physiology of the cell's cilia. There is just too much to know nowadays for any one person (or even lab group) to be an expert on all of cell biology or cancer genetics or evolutionary biology. You will find nowhere near this level of specialization in psychology.
Thus, there seem to be two patterns of specialization in science, both of which increase as scientific fields mature. Because more mature fields develop sprawling research literatures, there is necessarily a division of cognitive labor in mastering these topics sufficiently to do novel work. This accounts for the exquisite (or agonizing) degree of "horizontal" or output/topic specialization, which leads scientists increasingly to specialize like farmers. Further, as fields mature and demonstrate their ability to produce knowledge and advance society's goals, increasingly many resources are directed toward them so that they can afford a higher degree of "vertical" or task specialization. Scientists in these mature fields thus begin to look increasingly like factory workers.
I say this without any value judgment. The division of physical labor has made the world rich and the division of cognitive labor has made the world smart. These patterns of specialization were not imposed from the top, but evolved organically through the market of ideas (and, let's not kid ourselves, equally through the market of money). But these trends in specialization are not without trade-offs. We should remember that the Public Understanding of Science folks could probably learn a thing or two from the Science Education folks, and vice versa. What correctives to this siloing will emerge over time, or whether there will be wholesale shifts in scientific specialization in light of technological advances, are anyone's guess.