During his senior year, he attended a class taught by Franz Boas concerning American Indian languages.

These were interrupted by service in the U.S. Army Signal Corps during World War II, for which he worked as a codebreaker in North Africa and participated with the landing at Casablanca.

In 1962, Greenberg relocated to the anthropology department at Stanford University in California, where he continued working for the rest of his life.

Like Noam Chomsky, Greenberg sought to discover the universal structures on which human language is based.

An argument to reconcile the Greenbergian and Chomskyan methods can be found in Linguistic Universals (2006), edited by Ricardo Mairal and Juana Gil.

Many who are strongly opposed to Greenberg's methods of language classification (see below) acknowledge the importance of his typological work.

Greenberg rejected the opinion, prevalent among linguists since the mid-20th century, that comparative reconstruction was the only method to discover relationships between languages.

He argued that borrowing, when it occurs, is concentrated in cultural vocabulary and clusters "in certain semantic areas", making it easy to detect (1957:39).

He believed that multilateral comparison was not in any way opposed to the comparative method, but is, on the contrary, its necessary first step (Greenberg, 1957:44).

According to him, comparative reconstruction should have the status of an explanatory theory for facts already established by language classification (Greenberg, 1957:45).

Most historical linguists (Campbell 2001:45) reject the use of mass comparison as a method for establishing genealogical relationships between languages.

Among the most outspoken critics of mass comparison have been Lyle Campbell, Donald Ringe, William Poser, and the late R. Larry Trask.

Their objection was methodological: if mass comparison is not a valid method, it cannot be expected to have brought order successfully out of the confusion of African languages.

[citation needed] However, the work of Stephen Wurm (1982) and Malcolm Ross (2005) has provided considerable evidence for his once-radical idea that these languages form a single genetic unit.

Wurm stated that the lexical similarities between Great Andamanese and the West Papuan and Timor–Alor families "are quite striking and amount to virtual formal identity [...] in a number of instances."

Specialists of the individual language families have found extensive inaccuracies and errors in Greenberg's data, such as including data from non-existent languages, erroneous transcriptions of the forms compared, misinterpretations of the meanings of words used for comparison, and entirely spurious forms.

[8][9][10][11][12][13] Historical linguists also reject the validity of the method of multilateral (or mass) comparison upon which the classification is based.

However, Harvard geneticist David Reich notes that recent genetic studies have identified patterns that support Greenberg's Amerind classification: the "First American” category.

Greenberg basically agreed with the Nostratic concept, though he stressed a deep internal division between its northern 'tier' (his Eurasiatic) and a southern 'tier' (principally Afroasiatic and Dravidian).

Similarly, Georgiy Starostin (2002) arrives at a tripartite overall grouping: he considers Afroasiatic, Nostratic and Elamite to be roughly equidistant and more closely related to each other than to any other language family.

His colleague and former student Merritt Ruhlen ensured the publication of the final volume of his Eurasiatic work (2002) after his death.