The History of the Sewing Machine

Just how long a tailor's list of tasks once was is preserved in an 1898 survey by the US Department of Labor. This investigation carefully documented, step by step, the entire sequence of operations required to make the same item by hand versus by machine. For a single suit of men's clothing, the process unfolded like this: sponging the cloth, rolling it, laying out patterns and marking, cutting, bundling the trimmings, sewing the seams, marking the positions for buttonholes and buttons, cutting the buttonholes, working those buttonholes, felling the lining to finish, pressing, sewing on the buttons, and finally inspecting and brushing. That is thirteen steps.

The starting point of this story is that this entire chain was once held in the hands of a single person. Looking at data from the same survey for producing 12 dozen shirts the old-fashioned way, the report noted that "in the production of hand-made shirts, one woman performed all the operations on each unit." It was the same story for four units of low-cost men's clothing. A single woman carried out every single operation needed to complete an entire garment on her own. Mastering that craft took years. A committee of the British House of Lords recorded that multiple witnesses expressed deep concern over the breakdown of the apprenticeship system in tailoring.

Yet we won't try to answer "Who invented the sewing machine?" here. There are simply too many moving parts to credit just one person. Two of the most widely known US patents are both titled "Improvement in Sewing-Machines"—both Howe's No. 4,750 in 1846 and Singer's No. 8,294 in 1851. Neither claimed to have invented the sewing machine itself; instead, they claimed five and three specific mechanisms, respectively. The first machine to actually operate on a real production floor belonged to Thimonnier, patented in France in 1830. According to Smithsonian records, when he was sewing army uniforms with 80 machines in a Paris workshop in 1841, a crowd of tailors, terrified that the invention would strip away their livelihood, stormed in and smashed the machines to pieces. Machine-breaking sits right near the opening chapter of this story. An 1860 newspaper article noted that the earliest year a sewing-related patent could be found in US records was 1842. And as a Smithsonian researcher wrote in 1968, the practical sewing machine did not spring from the solitary genius of one person, but accumulated across a century of trials, failures, and partial successes.

The U.S. Census Bureau captured this process in the industry's own words in its 1900 report on clothing. After the sewing machine came into widespread use, ready-made clothing was mainly produced by journeyman tailors, and the Census noted: "In those days, each tailor was a skilled workman and made the entire garment."

The next step took place at home. As German immigrants entered ready-made garment manufacturing, they introduced a division of labor, and the stage for that division was each family's home. The father ran the machine, while the mother and children handled basting, finishing, and buttonholes. This meant that the first place where a single garment began passing through multiple pairs of hands was not a factory, but a family home.

Next came the "task system," which took root in New York around 1876. The wording in the Census report is decisive: "The introduction of this system revolutionized the ready-made clothing industry. Workers no longer needed to be skilled artisans, and journeyman tailors were pushed out of this branch." However, there was a qualification: "This system was applied only to the manufacture of coats." Trousers and vests were not yet included. The division of labor arrived at different speeds for different types of garments.

A team under the task system consisted of three people: an operator who did all the machine sewing, a baster who fitted the shell and lining together and pressed the seams, and a finisher who secured the edges to be stitched and basted the armholes. Outside this team was a presser, plus a woman who sewed on buttons and made buttonholes. That made five. The figures recorded by the Census were striking: around the time the task system began, a journeyman tailor made an entire coat alone and earned $2.50 to $3, but in a task shop, five people made that same coat for $1.50 to $2.

It was not just the labor that was broken down into pieces. The unit of pricing also changed along with it. A tailor used to be paid a set amount per coat. A team under the task system was paid per day's task. This difference came to define the story of the next several decades.

The description in the 1900 census is specific. When the task system first began, one task of overcoats was about 50 coats per week; the operator received $18, the baster $16, the finisher $11, and the presser $12. The woman handling buttons and buttonholes received a weekly wage of about $4.50. But when the economy slowed and labor was plentiful, manufacturers demanded lower prices from contractors. The contractors passed those terms down to the workers, and the workers agreed to increase the size of a task so that their pay would not be cut. A task that was initially 8 or 9 coats a day climbed up to 20 coats.

The next sentence recorded by the census captures the core of this entire chapter: "Because the workers always received the same pay whether a task was eight coats or twenty, they did not consider their wages to have been cut." When they could no longer finish a day's task in 12 hours a day, they worked longer. The report notes that tasks were often not finished even after 20 hours, frequently taking more than 30 hours.

In 1911, a study submitted to the Senate by the U.S. Bureau of Labor summed up this dynamic in a single line: wages under the task system were "in name a day wage, but in reality piecework." With no fixed hours, the report noted, workers felt compelled to fight the clock to complete the task, while employers were tempted to keep expanding its size. The unit of pricing was precisely what decided who held the advantage. Later, as production shifted to large factories, this unit changed once again to a weekly wage.

There is a sentence that is still quoted everywhere today: that making a shirt by hand takes 14 hours, while a sewing machine can do it in 1 hour and 16 minutes. You can even find this line on the history pages of sewing machine companies, often side by side with the claim that “in those days, 400 sewing machines could do whatever work 2,000 workers could do.”

I tracked down the original table. Quoted by a Smithsonian researcher citing an 1861 publication, it turned out to be an experiment conducted by the sewing machine company Wheeler & Wilson itself. The sample consisted of four hand sewers and four machine operators, and the figures were their averages. What is more, the title of the table was “Time of Stitching Garments.” What was measured was not the time it took to make a garment, but the time spent stitching. It did not include cutting, basting, making buttonholes, or pressing. A newspaper article from January 7, 1860, a year earlier, carried a very similar phrasing: “The stitching on a gentleman’s shirt can be easily done in one hour with a first-class sewing machine, while doing the same work by hand takes fourteen hours.” It was “the stitching,” not the time to make the shirt. The original source for the numbers 400 machines and 2,000 workers is even more surprising: it was a factory owner's own testimony presented in a legal petition arguing to extend an inventor's patent.

Then what about the numbers measured by the government? In an 1898 survey by the U.S. Department of Labor measuring in units of 12 dozen shirts, one unit showed time dropping from 720 hours to 83 hours and 24 minutes—a ninefold increase in speed—while another unit went from 1,176 hours by hand to 168 hours by machine, a sevenfold increase. A fourteenfold increase was nowhere to be found.

Yet this investigation also stands on shaky ground. And the report itself openly noted where it wobbles. The report's introduction explained that obtaining data on hand methods was far more difficult, noting that “it became necessary to seek out employers and workmen long since retired from active business, and from them elicit the necessary facts.” Even the years when the hand labor took place varied from unit to unit. While all machine data came from 1895, the hand data dated back to 1894, 1870, and 1853. There is an even more decisive flaw: the ratios diverged between two units of the exact same product. For the same 12 dozen cotton shirts, both took 720 hours by hand, but by machine one took 83 hours and 24 minutes while the other took 120 hours and 14.1 minutes. The ratio split between ninefold and sixfold. But we cannot simply say, “They were just different factories.” Looking at the report, the years the hand labor took place for these two units differed by 24 years—1870 and 1894. This means that even for identical products, this survey alone cannot tell us what caused the ratios to diverge.

Then what remains firm in this survey? It is the number of people. And the sentence in the report's introduction describing the factory method points right to it. It did not talk about speed: “an article passing successively from hand to hand, each workman performing a single operation, usually a very small part of the labor necessary to complete the article.”

An investigation submitted to the Senate by the U.S. Bureau of Labor in 1911 counted this number workshop by workshop. The investigators placed two ways of making a coat side by side and defined them.

One was the task system, which we saw in the previous chapter. The report described the task system as characterized by "employing a few skilled persons, each performing a bundle of several processes." Because of this, every member of the team needed all-around training and a broad knowledge of their branch. Under this method, a single coat passed through the hands of five or six people.

The other was the "section system." The report defined it this way: "Directly opposite the task system stands the section system. Here, work is subdivided into minute details, and each worker is restricted to a single process." It noted that this method was possible only in shops employing at least 90 to 100 people, and usually two or three times that number. It was a method that could work only with sufficient scale. In fact, an investigation of one contract shop included this line: "The work is carried on under a high division of labor. A coat passes through the hands of about 30 people before completion, compared to five or six in a task shop." That shop turned out around 800 coats per week, and the contractor estimated that coats could be made under the section system at half the cost of the task system, though it required an output of at least 900 coats per week to be viable.

Where manufacturers made garments directly in their own factories, the division was even finer. The report noted, "In some shops, a coat passes through the hands of sixty people before it is finished." It varied depending on the garment. While vests passed through 40 to 50 people in some places, dividing the work among only seven or eight people was also common; trousers were most commonly divided into six processes or fewer, though some shops broke them down into as many as sixty processes.

The 1900 Census pointed in the same direction: "Where a task shop would employ 20 people, a factory will employ 200." And it added that there were factories where over 100 people worked on a single coat.

This is the very heart of this chapter: what was broken down was not the work, but skill itself.

A single sentence from the 1900 US Census captures this: “Each worker's task is so limited that they soon become skilled and remarkably fast, but at the same time they are unable to make any other part of the garment.” The first clause and the second clause sit inside the exact same sentence. Because they perform only one motion, they become exceptionally good at it; and because they perform only one motion, they cannot do the rest. The census report went on to note that while this specialization allowed workers to learn their jobs in a short time, it also placed them completely under the manufacturer's control. It explained that a manufacturer is far more independent of someone who makes only a small fraction of a garment than of someone who makes the whole piece.

Testimony recorded by a British House of Lords committee in 1890 is even more specific. Tailoring work was split into numerous divisions, creating foremen or cutters, basters, machinists, fellers, buttonhole makers, pressers, and general hands—with one witness stating it was divided into twenty-five branches. Then comes this sentence: “The division of labor is carried so far that a person who can press a coat does not necessarily know how to press a vest, and a vest presser is just as unqualified to press trousers.” Even though it was all pressing, once the garment changed, they could no longer do it.

Job titles themselves became the names of physical motions. That was true of the titles recorded in the 1900 US Census describing a collar factory. The person stamping the brand and size was a “stamper,” the one pasting parts together was a “paster,” the one stitching the top and sides was a “runner,” the one turning it inside out was a “turner,” and the one doing the final stitching was a “stitcher.” The report noted that turning required a certain degree of skill and experience, so rather than factory work, it was mostly done by people working from home.

And there was no turning back from this direction. A passage in the British House of Lords report notes that while it once took years to learn the trade, now that each person handled only a tiny fragment, it could be learned in a short time. Saying that the time required to learn had grown shorter was simply another way of saying that the scope of what was learned had narrowed.

"Machines replaced humans" is what people most often say at this point. Yet the primary sources point in the opposite direction. The number of people touching the same garment actually increased. What used to take one person for 12 dozen shirts grew to 14, 22, 38, and 230, and what took five or six people for a single coat became 30 and 60.

There is one place where the numbers seem to shrink. Wage earners in U.S. men's clothing factories dropped from 144,926 in 1890 to 120,950 in 1900. Yet over the same period, the number of establishments actually increased from 4,867 to 5,731. Explaining this decrease, the census noted that part of the wages had shifted to contract work categorized under miscellaneous expenses. And right at the beginning of the same report, there is a confession: enumerators could not obtain information from many places where garment making took place, most of which were small workshops behind tenements and living quarters. It also noted that it was practically impossible to determine how many women were working from home. So subtracting these two numbers to calculate "how many jobs were lost" gives the wrong answer. What decreased was not the people, but the people who were counted.

There is one more surprising fact. Sewing machines of this era were mostly foot-treadle machines. Out of 5,731 U.S. men's clothing factories in 1900, only 1,519, or 26.5%, used mechanical power. In 1890, that figure was 8.3%. The census noted that in New York, mechanical power was used only in the largest factories, and even in buildings entirely occupied by garment workshops, the power to run the machines was often supplied by the operators themselves. This means three out of four ran without external power. What divided the labor was not mechanical power, but organization.

Other stories unfolded during this period as well. How needlework for hire was once work done outside the home, only to return inside with the domestic sewing machine; the still-unsettled debate over whether "the sewing machine liberated women"; and the story of displaced journeyman tailors. All three are true, and all three deserve to be told at length. But the central thread of this chapter is elsewhere: the scope of what a single person knows.

We see the very same division of labor in Korea, and we can look at how that division is recorded in official statistical documents today.

In the appendix of the "Survey on the Status of Garment Manufacturers"—a nationally approved statistical survey—there is an occupational classification. Designers, pattern makers, sample makers, cutters, sewing machine operators (sewers), linking operators, and production managers each have their own separate entry. However, the "sida" (assistant) is not an independent job category. It appears only as an example in the description for "Other textile-related craft workers," which lists "pressers (ironing sample products), inspectors, assistants (sida), etc." In contrast, the questionnaire from a 2025 survey published by the Seoul Workers' Rights Center lists cutters, sewing machine operators, sida, and siage (finishers) side by side as options. Two different official documents count the exact same position differently.

The description in that Seoul survey aligns directly with the core theme of this chapter. The report notes: "Garment workers usually handle only 1 to 2 processes repeatedly, and the hierarchical career path of growing into skilled workers by starting as a sida or assistant is still maintained." Because this sentence cites prior research, it is best understood as the report summarizing existing literature. Looking at the actual responses, out of 359 workers, 218 were sewing machine operators (60.7%), 52 were cutters (14.5%), and 42 were sida (11.7%). Furthermore, 86.8% of the respondents belonged to small workplaces with 1 to 5 employees.

The unit used to price labor also appears in this report. The same report defines "gaekgong" (freelance pieceworkers) as an employment arrangement where workers are not affiliated with a specific workplace and are paid in proportion to a set unit price per piece and the volume of finished work. In fact, 34.5% of respondents worked under other employment arrangements, including gaekgong. The same report records two contrasting sides of this practice: on one hand, a statement that business owners can cut fixed costs, making it mutually beneficial for both the gaekgong team and the employer (a point also cited from prior research); on the other hand, an interview quote stating, "Seriously, I was paid 1,000 won (about $0.75) 10 years ago, and it's still the exact same today. Piece rates just don't go up." Both statements exist within the very same report.

The scale of these shops is exceedingly small. As of 2021, nationwide garment manufacturers numbered 29,807, with 120,999 workers, averaging 4.2 people per establishment. In Seoul, 85.9% of the surveyed businesses had 4 or fewer workers, and in Jongno-gu, it was 97.3%. Moreover, 43.7% of workers were in their 50s, and 36.2% were 60 or older.

There is a passage in the 1900 US Census: "A coat may be made in New York City, the vest in Williamsburg, and the trousers in Newark, and the different garments brought together into a suit when they reach the wholesaler." It also notes that contractors usually made only one item—coats if coats, trousers if trousers—and that finding a workshop that made an entire suit was exceptionally rare.

Today, that exact same structure exists not across cities, but across countries. In a 2025 brief, the International Labour Organization noted that the garment industry alone employs more than 90 million people worldwide, with 75% of them in Asia.

The division within the factory floor remains as well. A 2017 sectoral brief published by the same organization described the breakdown of garment factory labor: sewing 67%, ironing and finishing 10%, samples and fabric inspection 10%, pattern making and cutting 8%, and trimming and spreading 5%. Two-thirds of the workforce is concentrated in a single process. And in that same brief, there is a sentence that directly echoes the central theme of this chapter: that while sewing machine operators do not require formal education, cutting machine operators require technical training. A line is still drawn today in the exact same place it was a hundred years ago in New York, when cutting stayed inside the manufacturer's building while everything else moved outside.

That said, it is worth clarifying that the brief's outlook on automation was a projection. Anticipating that manual sewing could be replaced as automated sewing machines emerged, the brief included an estimate that a large share of garment workers in Vietnam and Cambodia were at risk of automation. We were unable to confirm through primary sources in this investigation how that forecast ultimately unfolded. That is why we record the projection here strictly as a projection.

Instead of making predictions, let me note down what to watch. There are four things.

First is the number of steps handled by one person. That is the thread this series has followed. It narrowed from an entire garment down to a bundle of multiple steps, and then down to a single process. Current surveys in Korea record that a garment worker usually handles 1–2 steps. Watching whether this number shrinks further or expands will show you the direction. An expansion would not be unprecedented; after all, in small workshops, one person has to handle multiple steps.

Second is the unit used to price labor. Is it per garment, a daily task, a weekly wage, or a piece rate per task? Every time this unit changed, the value of the same labor changed with it. Watch which unit Korea's *gaekgong* (freelance pieceworkers) stand on today, and whether that unit shifts.

Third is which process is handed over to machines next. Cutting was handed over to machines long ago, and so were buttonholes. Sewing still relies heavily on human hands. The fact that two-thirds of the workforce is concentrated in sewing, according to data from international organizations, is a clear sign. Seeing whether this task is handed over, and where those people go if it is, will show you the next chapter of this story.

Fourth is the way we count. Over a hundred years ago, U.S. statistical agencies noted themselves that they failed to count work done in rented tenement rooms. Today, Korea's official government-approved statistics do not count *sida* (sewing assistants) as an independent occupation. What you count determines what you see. Watch to see if the classifications change.

There is one thing this series has left undecided to the very end: whether the sewing machine did a good thing or a bad thing. What we confirmed instead is this: What disappeared was not the person who sews, but the state of one person knowing a whole garment from start to finish. The number of sewing workers actually grew; it was only the scope of what each person knew that shrank.