This paper develops a corpus-linguistic approach to the demography of North American cities. In a groundbreaking study, Chomsky (1957:17) convincingly showed that it can be proven on linguistic grounds alone that more people live in New York than in Dayton, Ohio. Unfortunately, Chomsky did not go on to develop his corpus methods any further.

In the present study, all occurrences of the search string "I live in" were extracted from the American National Corpus. The word forms immediately to the right of the search string were extracted from the concordance and ranked according to frequency. Table 1 shows the ten items with the highest token frequencies:

The table reveals a number of surprising results. Besides previously unknown cities, also a number of cities previously believed to lie elsewhere are actually found in North America. It will be up to geographers to face the challenge and actually put these on the American map.

To obtain demographic information about these cities, the token numbers must be related to the numbers of inhabitants. To calculate these figures, it is necessary to know the exact number of inhabitants of one city. This number, divided by its token frequency, gives us the token-inhabitant ratio that we need. Chomsky (personal communication) gives an estimate of the population of Dayton, Ohio with 166,179. Divided by its token frequency 88, the corpus-linguistic demographical constant (CLDC) is exactly

With this figure in mind, I believe that corpus-based armchair demography has a number of advantages over more traditional methods of statistical demography. Not only is it fast and economical, it also points toward new phenomena that traditional demography did not have any account for.

	Poetry Corner
	Phonemic Color--Tong Shunming
	SpecGram Vol CL, No 1 Contents