From Miniskirts to Midi Lengths Fashion Keeps Returning Every Two Decades Even as Trends Fragment, Study Finds

Northwestern University researchers have developed a mathematical model showing that fashion trends follow recurring cycles of roughly 20 years, based on analysis of around 37,000 images of women’s clothing spanning more than a century. The study provided quantitative backing to a long-cited industry observation and indicated that trend patterns have become more fragmented in recent decades.

Long Story, Cut Short
  • A Northwestern University study analysing around 37,000 images found that fashion trends tend to return in cycles of about 20 years.
  • Researchers measured dress features and modelled how the tension between conformity and differentiation drives styles to rise, fade and return.
  • The study also found that fashion has become more fragmented since the 1980s, with multiple skirt lengths appearing at the same time.
Representative dresses from 1923 to 1987, showing the change in hemlines over time.
Hemline Changes Representative dresses from 1923 to 1987, showing the change in hemlines over time. Emma Zajdela / Daniel Abrams / Commercial Pattern Archive

Fashion trends follow recurring cycles of roughly 20 years, with styles rising, fading and returning in a measurable and predictable pattern over time, a mathematical model based on more than a century of clothing data shows. The pattern is driven by a balance between differentiation and conformity in design choices. The findings also indicate that this cyclical structure has weakened in recent decades as fashion becomes more fragmented.

  • The analysis tracked approximately 37,000 images of women’s clothing spanning from the late nineteenth century to the present, forming one of the most extensive datasets of its kind.
  • Researchers quantified features such as hemlines, necklines and waistlines to convert garments into measurable data across decades for consistent comparison.
  • The model captures how designers shift away from overly common styles while remaining within acceptable boundaries of wearability.
  • The findings were presented recently at the American Physical Society Global Physics Summit in Denver by lead author Emma Zajdela.

BUILDING THE MODEL: A mathematical model based on one of the largest quantitative fashion datasets has been developed to examine long-term style dynamics across more than a century. The dataset combined historical sewing patterns and runway collections, enabling systematic measurement of garment features over time. Researchers translated design elements into numerical values, allowing fashion trends to be tracked, compared and analysed as a measurable system across decades with consistent parameters.

  • The dataset drew on the Commercial Pattern Archive at the University of Rhode Island alongside runway collections spanning from the late nineteenth century to the present.
  • Researchers analysed a large set of images of women’s clothing to build a consistent, long-term dataset of fashion features.
  • Key garment attributes were measured using custom tools to enable quantitative comparison of design features across decades.
  • Emma Zajdela conducted the work as a PhD candidate at Northwestern’s McCormick School of Engineering under Daniel Abrams, professor of engineering sciences and applied mathematics and co-director of the Northwestern Institute on Complex Systems, with co-authors Alicia Caticha of Northwestern’s Weinberg College of Arts and Sciences, and Jeremy White and Emily Kohlberg of Abrams’ research group.

CYCLE IN DATA: Fashion styles rise, fall and return in a repeating pattern that peaks roughly every two decades, forming a measurable cycle across long-term clothing data. The pattern appears as a wave, with gradual shifts punctuated by periodic reversals in dominant styles. One of the clearest examples is hemline length, which has oscillated between shorter and longer forms across successive generations.

  • Skirt lengths moved from shorter flapper styles in the 1920s to longer, more conservative designs in the 1950s and then to miniskirts in the late 1960s.
  • The cyclical pattern reflects how styles gain popularity, decline as they become widespread, and later re-emerge in modified forms.
  • The repeating wave structure was observed consistently across more than a century of clothing data analysed in the study.
  • The findings align with the long-cited industry observation that fashion trends tend to return approximately every 20 years.

SHIFT IN PATTERNS: The clear cyclical structure observed across much of the twentieth century has become less distinct in recent decades, as fashion trends increasingly overlap and diverge. Instead of a single dominant style rising and falling in a predictable wave, multiple trends now coexist, reducing the sharpness of cyclical peaks and troughs. This shift reflects a broader fragmentation in fashion, with greater variation and multiple styles coexisting within the same periods.

  • Since the late twentieth century, skirt lengths and other design features have shown greater variation within the same time periods.
  • The data indicates that no single style consistently dominates, with shorter and longer hemlines appearing simultaneously.
  • The weakening of clear cycles suggests that conformity pressures have reduced relative to earlier periods.
  • The findings indicate that fashion trends have become more fragmented, with a wider range of styles appearing simultaneously.
Representative dresses from 1920 to 2010.
Changes over Time Representative dresses from 1920 to 2010. Emma Zajdela / Daniel Abrams / Commercial Pattern Archive

BEYOND FASHION: The cyclical dynamics observed in fashion reflect wider patterns in how ideas and behaviours spread across social systems over extended periods. The model shows that trends emerge and recede through a balance between conformity and differentiation, a mechanism that extends beyond clothing into other domains where collective behaviour evolves over time.

  • The findings suggest that similar cyclical patterns may be present in how ideas spread across society, where collective preferences rise, peak and decline over time.
  • The balance between fitting in and standing out acts as a general driver of change in complex social systems, shaping how individuals respond to prevailing trends.
  • Researchers show that fashion data can be used to study how trends propagate, interact and transform across large populations over extended timeframes.
  • The study contributes to a broader body of work applying statistical physics approaches to understand how patterns of behaviour emerge and evolve in society.

WHAT THEY SAID

To our knowledge, this is the first time that someone developed such an extensive and precise database of fashion measures across more than a century. ... In more recent years, there are more options: really short dresses, floor-length dresses and midi dresses. There is an increase in variance over time and less conformity.

Emma Zajdela
Postdoctoral Fellow and Research Fellow
Princeton University and Santa Fe Institute

Fashion is a cultural phenomenon, so the way it changes over time reflects how people behave. ... Over time, this constant push to be different from the recent past causes styles to swing back and forth. The system intrinsically wants to oscillate, and we see those cycles in the data.

Daniel Abrams
Professor of Engineering Sciences and Applied Mathematics, Co-Director of the Northwestern Institute on Complex Systems
Northwestern University

 
 
Dated posted: 19 March 2026 Last modified: 19 March 2026