Mid-career pitfall of consecutive success in science

Abstract

The creativity of scientists often manifests as localized hot streaks of significant success. Understanding the underlying mechanisms of these influential phases can enhance the effectiveness of support systems and funding allocation, fostering groundbreaking discoveries worthy of accolades. Historically, analyses have suggested that hot streaks occur randomly over time. However, our research, through meticulous examination, reveals that these phases are not flatly distributed but are more frequent at the early and late stages of scientists' careers. Notably, both early and late hot streaks are marked by dense tie collaborations, with the former typically involving close partnerships with particular authors and the latter being characterized by involvement in large-scale projects compared with single-top or ordinary papers. This pattern indicates that mid-career researchers lack both intimate relations and resources to keep big projects, leading to``mid-career pitfal'' of consecutive success. This insight holds profound implications for the development of policies and initiatives aimed at bolstering innovative research and discovery.

Figures (20)

• Figure 1: (a) Definition of consecutive success. Top $k$% impactful papers are shown as blue circles and they appear $X$ times within $N$ publications with parameters $X=3, N=5, k=10\%$. We use the number of publications in sequence as the length of the researchers’ career instead of years. (b-d) Probability distribution when consecutive success occurs in the scientists' careers. The x-axis, relative timing, represents normalized career length. The numbers in parentheses represent the percentages of researchers, out of 100,000, who have experienced at least one hot streak under that parameter. As consecutive success becomes stronger, the number of researchers experiencing consecutive success decreases. (b) The distributions of consecutive success with different length-related parameters $X$ and $N$ ($k=10\%$). A single success ($X/N=1/1$) occurs with a constant probability but long continuous successes ($X/N=5/9$) occur more frequently in the early and late throughout the career. (c) The distributions of consecutive success for different $k$ ($X/N=3/5$). Big success occurs more in the early and late stages. (d) compares the distribution between the raw and shuffled career sequences with parameters $X/N=3/5$ and $k=10\%$.
• Figure 2: Characteristics of consecutive success. (a) A typical example of a 'Hot' career sequence. The period filled in orange is detected as a hot streak of five consecutive papers. Blue dots represent top 10% papers, appearing three times within the five papers. (b) An example of 'Top' sequence, represents a period of single success, not consecutive. A green dot indicates a top 10% paper, appearing once among five papers. (c) 'Ordinary' sequence example as comparisons, five consecutive non-top 10% papers, representing a period without significant hits in careers. (d-i) Comparison of the 3 sequences across 6 metrics. The relative career timing is divided into ten bins, and metrics are calculated for sequences with the same start timing in each bin. This allows observation of how metrics vary in early, mid, and late career stages. (d) Team size, focused on author lists of fewer than 10 people (Supplementary information \ref{['sec_sup:team']}). (e) Proportion of teams with 10+ members. (f) Proportion of sequences with dense ties, appearing 3 or more times in the co-author list of the 5 consecutive papers. (g) Betweenness centrality of the focal author in the co-authorship network of the 5 papers. (h) The number of topics tackled in the 5 papers, ranging from 1 to 5. (i) Proportion of new topics not previously tackled. In (d) and (f-i), calculations are performed on sequences where the team size for all 5 papers is less than 10. The bars in (d,g,h) represent the 95% confidence interval that includes the estimated population mean, assuming a normal distribution. See \ref{['sec:methods']} for details of definitions and calculations of each metric.
• Figure 3: Four types of consecutive successes. (a) The histogram when consecutive successes occur with parameters $X/N=3/5$ and $k=10\%$. The four types of them are displayed in a stacked manner, each represented by a different color. The top two types, shown in pink and red, represent consecutive successes with dense ties, more common in early and late stages. The bottom two, in light blue and blue, depict loose ties occurring consistently across the career. (b) The four types' proportions. 65% of them originated from dense ties and 40% originated from large teams. (c-f) Typical co-authorship networks for each type. Green nodes represent the focal author, other nodes are co-authors, and edges indicate the number of collaborations in five consecutive papers, ranging from one to five. (c) Dense ties in small teams are shown as thick edges between a few nodes. (d) Thick edges with large teams composed of many co-authors. (e) The focal author is connected with fewer authors and thinner edges representing loose ties and small teams. (f) No thick edges with many co-authors show collaboration in loose ties with large teams.
• Figure 4: Two scientists' typical career dynamics. (a,b) Career sequences. The papers with top-10% impact are colored and the orange areas indicate consecutive success identified by our method. (c) Relation between papers' publication timing in normalized career length and their team size for two scientists. (d) Frequency of co-authors in five consecutive papers; higher values mean repeated collaborations with the same authors, reflecting dense ties. (e) Topic diversity dependency on career stages, measured by the variety of topics in five consecutive papers. For clarity, (c-e) display moving averages with a window size of 5% career length.
• Figure S1: The distribution of consecutive success among researchers with varying career lengths in five-year increments. Regardless of age, the probability of consecutive success declines from seven years after the career begins to 5-10 years before the career ends.