The able research assistance of John Nader and Martin Kohli is gratefully acknowledged, as are the helpful comments on earlier drafts by Robert Heilbroner, David Gordon, Claudia Goldin, and Floria Thomson.Google Scholar

¹ This notion of technological change is developed in Thomson, Ross, “Technological Change As New Product Development,” Social Concept, 3 (06 1986), pp. 3–26.Google Scholar On new product development see Levine, David, Economic Theory (London, 1978), vol. 2, chaps. 4, 5;Google Scholar and Shapiro, Nina, “Innovation, New Industries and New Firms,” Eastern Economic Journal, 12 (01–03 1986), pp. 27–43.Google Scholar On the role of learning in technological change, see David, Paul, Technical Choice, Innovation and Economic Growth (Cambridge, 1975).Google Scholar

² The early development of the sewing machine also depended upon communication through attempts to sell; Howe's machine only developed to practicality through these means. I take this up in a paper, “Between Invention and Practicality: The Development of the Sewing Machine” (unpublished manuscript, New School for Social Research, 1986) and in a book manuscript, “The Birth of Mechanized Production in the United States” (unpublished manuscript, New School for Social Research, 1987).Google Scholar

³ While available, records of the sewing machine patent poo1 are incomplete especially before 1865. They include only about half of the sewing machines recorded in the 1860 manufacturing census.Google Scholar

⁴ The series of agencies excludes outlets which specialized in sewing machine repair and attachments, though these are not always indicated as such. It also misses some agencies, particularly in earlier years when many directories had lists of advertisers but no separate business directories categorized by type of product.Google Scholar

⁵ The issue of using patents to measure invention is carefully discussed in Schmookler, Jacob, Invention and Economic Growth (Cambridge, Mass., 1966)CrossRefGoogle Scholar. I agree with his argument that patents usefully measure inventive activity, as opposed to the effect of this activity on productivity. The list of sewing machine patents was taken from class 112, U. S. Department of Commerce, Patent and Trademark Office, U.S. Patent Classification—Subclass Listing (30th edn., Washington D. C., 1985). Patents were then individually consulted to form the data base.Google Scholar

⁶ Patents issued to more than one person were classified as initial if none of the patentees had prior sewing machine patents but at least one had later patents, as repeats if at least one patentee had prior patents, and as single if none of the patentees had other sewing machine patents.Google Scholar

⁷ This lag represents 18 months after purchase to perceive and solve technical problems and 6 months between patent application and issuance. In some cases, it seems that a longer lag or a lag distributed over several years would be more appropriate, but much the same results would be obtained.Google Scholar

⁸ Note that the ratio of first patents to agencies declined over time and first patent-sales ratio fell even more rapidly. Especially for special-purpose machines, this decline may have resulted from a narrowing of inventors to those with more technological knowledge. Furthermore, as more were involved, the likelihood of redundant invention increased. A different specification of the learning thesis might also account for this decline. Relevant learning surely occurred on machines over two years old, if not by the buyer then by other family members, repairers, and second-hand dealers. To grasp this more protracted learning, first patents might be related to cumulated sales net of abandoned machines. With such a measure, growth of machines in use would in general decelerate as the penetration process proceeded and as such would more closely match the first patent trend. This approach, however, cannot readily explain the fall in first patents in the mid-1870s.Google Scholar

⁹ These conclusions also hold within categories of states. Southern tertiary states regularly had fewer first patents and agencies per capita than northern and western tertiary states. Likewise, among nodal states New York consistently had the fewest first patents and agencies per capita.Google Scholar

¹⁰ Companies with more outside agencies would of course be expected to learn more than those with fewer. To weigh the number of companies by the number of their agencies outside the state (or category of states) in any period would accentuate the importance of nodal states. The geographic concentration of production, a consequence of the growth of interstate agency systems, may also reflect the significance of interstate learning. Nodal states had by far the largest per capita employment in sewing machine manufacturing. From 1860 through 1880, they averaged 610 per million population, compared with 187 for secondary states and only 4 for tertiary states.Google Scholar

¹¹ Use of the sewing machine as a producer good may also have accounted for the higher first patent-agency ratios in nodal states by posing distinct problems involving more intense interaction with sewing machine producers, and therefore inducing more patenting. This interpretation is supported by the fact that the shoe and clothing industries were concentrated in nodal states and that the share of patents which could not be used on family machines, though hard to ascertain, was surely much higher than the share of manufacturing machines in total sales.Google Scholar

¹² Table 5 differs from earlier tables because it focuses on inventors rather than patents; a patent issued to several individuals is counted separately for each patentee here but was counted only once in earlier tables. It also includes the 26 patents issued prior to 1853.Google Scholar

¹³ This notion of use only roughly approximates actual usage. It understates use because it takes little account of patents assigned after the issuance of the patent or of licensing, yet both were common. (For an interesting example of the fruits and difficulties of locating such assignments, see Cooper's, Carolyn C. dissertation “The Roles of Thomas Blanchard's Woodworking Inventions in 19th Century American Manufacturing Technology” Yale University, 1985, summarized in this issue of the JOURNAL).Google Scholar The sources are less adequate concerning the usage of attachments and special-purpose machines and of patents outside surveyed cities, though the conclusions of this study would stand if only patents for general-purpose machines in surveyed cities are considered. Since the category used patents refers to all patents issued to inventors with at least one used patent, it surely overstates actual usage. But even patents which ultimately failed could occasion interactions which stimulated further inventing, especially when inventors owned sewing machine firms or regularly assigned to such firms.

¹⁴ Use and learning also varied by type of machine. I consider this issue for shoe versus other sewing machines in my manuscript “The Birth of Mechanized Production in the United States.”Google Scholar

¹⁵ When classified by kind of use, inventors varied little in the frequency of repeat patenting; virtually all the difference of repeat patenting rates is accounted for by variation among multiple inventors in the number of repeats. This contrasts with unused patents, for which the far lower frequency of repeat inventing accounts for somewhat over half of the lower rate. Perhaps use of any kind fostered repeat inventing, but the incentive and finance provided by larger-scale usage was decisive in the greater persistence of inventors for larger firms.Google Scholar

¹⁶ Maturity does not mean that invention ceased, as the significant upturn in patenting in 1881 and 1882 indicates. Still, this upturn may have been a response to the expiration of another institution formed early in the product cycle, the patent pool.Google Scholar