DURING THE EARLY 1980s, Daniel was exploring the social universe of San Miguel Bay fishers, but his main obsession was with something else entirely. Since his initiatory journey in Indonesia, he had been searching for new techniques for determining the growth rate of the numerous species of fish caught by tropical fisheries. The statistical methods developed by scientists from temperate climates wouldn’t work because they required that researchers already know the age of the fish in each size class. Daniel dug through the literature, armed with a clipboard and a pencil that he sharpened with an electric pencil sharpener, his favorite gadget at the time. A second pencil was usually stuck behind his ear in case he lost the first one. At ICLARM, he already had a solid reputation as a mad genius, often walking around barefoot or in his socks among his white-collared colleagues, jubilantly ignoring administrative constraints, and stirring his coffee with his coworkers’ pens.
During one of his bibliographic orgies, Daniel discovered that the Danes, before becoming world-class specialists in determining age from otolith cross-sections, had developed much simpler methods. In particular, toward the end of the nineteenth century, C. G. Johannes Petersen, a young marine biologist at the University of Copenhagen, suggested finding the growth rate of fish by examining the size classes of individuals of the same species captured over time.
To visualize this method, let’s imagine a large group of schoolchildren that includes every grade from kindergarten to the end of high school. In this slightly abnormal human population, children are born during one or two brief periods of the year, and groups of different sizes appear very clearly: we call these “cohorts.” So, sixth graders from a spring cohort would be slightly larger, on average, than their classmates from the autumn cohort. Each year, the school nurses measure the children’s height, but the medical service is very disorganized and writes down neither the date the measurements were taken nor the name of the student or their grade level. Using this anonymous data, though, we can still count the number of students in each size class. For a given year, there will be more students in size classes that correspond to cohorts than for size classes that fall between cohorts. If we were to represent this annual tally as a diagram,* we would see a series of small hills representing more numerous size classes. Then, we can align the diagrams from several years on a piece of paper, one per line, with the oldest at the top and the most recent at the bottom.
Looking over this page, we could see how the small hills move toward the right from line to line: the large cohort of sixth graders born in the spring would have grown a few inches when they were measured in seventh grade, then in eighth. Growth is especially visible among younger individuals, so we needn’t continue our measuring program after they leave for college. In any case, when the size of a given cohort stops progressing, this gives us the size of the average adult for a group of students, and the time it took them to reach their adult size allows us to calculate the mean annual growth rate.
In the nineteenth century, without computers, statistical methods were primitive: C. G. Johannes Petersen, who was interested in the growth of fish, not schoolchildren, filled up entire pages with hills and valleys obtained by measuring thousands of anonymous fish (of the same species) captured over the years in the same area. After that, he drew diagonal lines with a ruler from the top left to the bottom right to connect the hills that supposedly corresponded to different cohorts. This is how he determined the maximal size and the growth rate using an approximative, visual method. Petersen’s approach was much used, though with some variation, until the middle of the twentieth century, when it fell out of favor because of its lack of precision. Daniel, who had resurrected von Bertalanffy, did the same for Petersen by making two major adjustments.
First, he developed a new method for “smoothing” the hills and valleys so that the height of the hills wouldn’t be impacted by the number of fish measured. “At the time, I mentally smoothed anything I saw—mountains on the horizon, women’s breasts . . . smoothing haunted me.”
Next, he developed a statistical approach that would allow him to calculate growth rate and maximal size by “trial and error.” First, Daniel used von Bertalanffy’s equation to determine the approximate growth parameters that were supposedly valid for the species of fish being studied. He then compared the supposed growth curve with the hill-and-valley diagram: the goal was to obtain a curve that passed through the maximum number of summits while avoiding the valleys. Little by little, he adjusted the parameters to obtain the curve closest to the ideal path. Daniel didn’t mention this explicitly in his work from the early 1980s, but the technique he used was Bayesian. Bayesian statistics* are fascinating because they are based on a huge amount of research about a notion that seems extremely unscientific: a priori assumptions (or “priors”). If we create somewhat of a caricature, we might define a Bayesian as “one who, vaguely expecting a horse and catching a glimpse of a donkey, strongly believes he has seen a mule.”† In the same way, Daniel had a priori knowledge of the maximal size and growth rate of the fish he was studying and used a likelihood principle to arrive at the solution that seemed closest to the truth. This process of statistical trial and error replaced and perfected the more approximate graphing methods by using formal criteria. In his analytical approach, Daniel was likely influenced by his times, as Bayesian statistics made a big comeback in the 1980s. These methods required a lot of computation, however.
“Luckily, computers replaced the visual line-drawing method; today we would call it artificial intelligence,” Daniel remembers. As had become his habit, he quickly forged an alliance with an expert—this time, a programmer, a Filipino named Noel David, who coded a computational routine in BASIC. The two colleagues named the program ELEFAN (for “Electronic Length Frequency Analysis”), with a wink and a nod that would become the ICLARM gang’s trademark during the 1980s. The first version of the program even featured a line of three stylized elephants walking across the opening screen. The routine initially ran on the first microcomputers that replaced the outdated calculators at the end of the 1970s. These “little” contraptions, like the TRS-80, looked like huge cathode ray tube television sets, with matching keyboards as thick as shoeboxes and maximum memory capacities of forty-eight kilobytes (a million times smaller than a basic computer today). At the controls was Jingles, who, driven from the San Miguel project by Antonio Mines, had become Daniel’s very first intern. “I already had some knowledge of BASIC, but FORTRAN was the ‘in’ thing at the time,” Jingles recalls. The two were thick as thieves. Inspired by Daniel, Jingles dove into the fisheries literature and began trawling for data. “All of it was in storage in the basement. I spent maybe three months digging up information in the Bureau of Fisheries, in the dust and dirt and sweat. I was actually rebuilding all these datasets. I loaded them species by species into ELEFAN, and it took hours before the first values came out . . . I was often working late at night.”
AT THE SAME time, Daniel taught Jingles the basics of fisheries biology: “Daniel’s big thing was independence; he really wanted to make sure that he gave me everything I needed to keep learning by myself.” This desire to help others to learn also led him to teach at the University of the Philippines, on top of his already-hectic schedule at ICLARM. He lived halfway between ICLARM and the Diliman campus, regularly zipping back and forth in traffic that was much lighter than the ubiquitous nightmare that characterizes Manila today. Daniel taught a single class: “Fish Population Dynamics,” but his Filipino students wouldn’t forget him easily. Jingles remembers Professor Pauly as “an alien, like a creature from another world totally obsessed with science, but who explained things well, without being condescending to his students.” “Sometimes he would make a cutting remark, for which he would apologize . . . I never saw him reject anybody who would ask for help,” Jingles adds, “and that is something that I noticed in my three, four years with him. The door of his office was always open, which is amazing for someone who wrote as much as he did.”
Over the years, ELEFAN would be completed by increasingly complex modules. It would allow researchers to estimate the size, and therefore the age, at which fish should be caught to ensure that they have had time to reproduce and maintain the population. This minimum could then help researchers to determine which mesh sizes can be used to fish sustainably. For Daniel, “it became possible to work on things that were totally impossible before in the tropical setting; it was outrageously simple.” In fact, the program used data that was easy to collect—just fish length measurements. These can be collected from a research vessel, but also in fish markets. When Pauly and David launched ELEFAN in the early 1980s, biology laboratories were already overflowing with length measurements stored up over several decades, and that data went to feed an increasingly hungry ELEFAN.
The program was such an immediate success, especially in developing countries, that Daniel and his colleagues soon found themselves traveling nonstop to host training sessions and work groups on the subject. But the criticism they received was also severe, especially from Western researchers. “Those guys came after me because they said it was too simple. A lot of people were against it. And then, after ten years, it became a normal thing,” Daniel recalls. Some early critics focused on the fact that fish growth rates vary seasonally, which can interfere with the analysis. In fact, even if seasonality is not especially noticeable in tropical waters, it does have a very real incidence on the growth of different cohorts of young fish. “So, we had to do all sorts of supplemental analyses and write a lot of articles, but gradually, we fixed the seasonal problem.”
What’s more, Daniel and his team quickly realized that their computational routine could be applied to almost all marine organisms, not just fish. In particular, ELEFAN would later be used a great deal on crustaceans, for which it is difficult to determine age classes, but also for a whole crowd of mollusks, turtles, marine mammals, even lizards. Jingles remembers making hundreds of copies of the program on floppy disks, the ancestor of the CD-ROM, and sending them to other researchers in developing countries. “Daniel’s passion, his obsession, was motivating his colleagues to attack the problem of tropical fish growth,” comments Jingles. John Gulland, the grand master of fish stock estimation at the time, put Daniel on his guard, however: “It’s not people using ELEFAN [properly] that I’m afraid of, it’s the misuse of it,” he said. In fact, as Jingles notes, “People [who don’t understand the program] use it like a cookbook, even today; it’s what goes in, and what goes out.”
In 1982, Jingles received a tempting offer from the GTZ that included the possibility of doing a master’s degree in the United Kingdom. Daniel was hurt, but quickly riposted with an offer of his own, complete with a program in Kiel. “I accepted because I could bring my girlfriend to Germany, and I thought that the weather in England was far colder. I was wrong on that second point!” laughs Jingles. He survived three years in Kiel, writing and defending his master’s thesis in German, the first of many foreign students whom Daniel would send to freeze on the coast of the Baltic Sea.
A brilliant career followed for Jingles, with jobs at the University of the Philippines, then the World Wildlife Fund, which he left in 2015. “Right now, I’m working on fish farming in the Philippines, just trying to show people that with aquaculture you can help enhance the environment rather than destroying it,” he says. “Here, the level of poverty is about 32 percent, but in the fisheries sector it’s about 42 percent, and the Philippine government isn’t doing anything to help.”
* Called a “frequency histogram.”
* Named for the British reverend and statistician Thomas Bayes (1701–61).
† Quote attributed to Karl Pearson.