Carnegie Mellon University
April 27, 2016

Solving the Puzzle of Language

By Emily Stimmel

When Edward (Ted) Gibson (DC’91) came to Carnegie Mellon University in 1986 to pursue a Ph.D. in computational linguistics, he saw language as a puzzle to be solved.

Fast-forward to today: Gibson is still putting the pieces together and making fascinating discoveries. And the world has taken notice. His research was recently featured in “Nature’s Numbers,” a BBC Discovery documentary series that examines the origins of mathematical understanding in humans.

That particular work started with a 2007 trip to Brazil to conduct cognitive psychology tests with the Pirahã people of the Amazon rainforest. He was invited by Daniel Everett, current dean of Arts and Sciences at Bentley University, whom he had met during his time at CMU.

Gibson and Everett noticed that the Pirahã could not perform tasks related to counting, and instead used relative values. They concluded that numbers must not be useful to the tribe because natural objects are not measured in discrete units. For example, fish aren’t identical in size; in order to equally divide resources to share, the villagers instead rely on a visual understanding of concepts like volume.

In the documentary, Gibson said, “Relative to the context, what might be ‘many’ could turn into ‘few.’”

A lack of numbers is one of several unusual features of the Pirahã language that violates supposedly universal linguistic principles. For instance, it has been proposed that the Pirahã language lacks recursion, a process through which sentences — and languages — can be repeatedly expanded. Gibson’s latest research on this topic is consistent with this hypothesis.

Gibson, professor of cognitive science at MIT, is currently using the Pirahã project as a springboard and assembling a corpus of texts from non-industrialized cultures, including the Tsimane’, an indigenous population in the Bolivian Amazon region.

“If we want to look at why human cognition is the way it is, we don’t want to restrict ourselves to industrialized populations,” Gibson remarked.

Gibson chose the Philosophy Department’s computational linguistics doctoral degree because it blended his interests in math, computer science and language.

“I just studied things that were fun for me,” said Gibson.

The interdisciplinary, positive environment of CMU encouraged him to incorporate cognitive psychology in his thesis — a computational model of the ways people understand sentences. Though psychology wasn’t a formal component of his Ph.D. program, Gibson regularly interacted with members of the department, including Psychology Professors Brian MacWhinney and David Plaut.

“During his years at CMU, he was already a stand-out and a potential star,” said MacWhinney. “He has uniquely good control over the three skills of computation, experimentation and linguistic analysis, which makes him able to produce top-rate psycholinguistics theory and research.”

The relationship between culture and cognition — particularly as it relates to language — is one of three areas that Gibson and his students investigate in MIT’s TedLab. They also study why human languages look the way they do and how people learn and process language.

Doug Rohde (SCS’02), a former graduate student of Plaut’s who now works at Google, joined TedLab as a postdoctoral fellow in order run empirical tests of a computational model he had developed in his Ph.D. work.

“Ted is nearly unique in his ability to integrate computational and empirical studies with formal linguistic analysis, and so working with Ted was ideal for Doug to extend his training and to develop intellectually," Plaut commented.

Gibson sees parallels between MIT and CMU, which he attributes to both universities placing value on quantitative work. He particularly appreciates the way that CMU integrates the humanities with a quantitative approach — an asset that MacWhinney also recognizes.

“Tools such as natural language processing and corpus analysis can be used to investigate issues near and dear to the humanities,” MacWhinney said.

Perhaps no subject is more innately linked to the humanities — and humankind — than language. To Gibson, this is what makes the work so appealing.

“The puzzle I most want to study is why language is the way it is. What makes a human especially human is language,” he said.