Ling 431/631: Corpus Linguistics

Ben Bergen

Meeting 6: Morphology

September 24, 2007

Productivity

One of the most interesting things about the grammar of human languages is the productivity of syntactic and morphological processes - namely the possibility of producing new things from component parts and processes. This is presumably the property that gives language its infinite potential.

But how do you measure it? (adapted loosely from Bauer 2001)

1. Productivity is how thoroughly a morphological process is distributed in the language:

P = n_type/H_type

where n_type is the number of types using the rule and H_type is the hypothetical total number of types which the word formation rule could give rise to.

Any problems with this?

2. Productivity is how likely a process is to be used in comparison with competitors

P = n_type/T_type or n_token/T_token

where n_type and n_token are the type and token frequencies of the process and T_type and T_token are the total type and token frequencies of the particular process and its competitors

Problems with these?

3. Productivity is how often a process is used to create new words, compared to its overall use

P = n₁/N_token

where n₁ is the number of words formed by the appropriate process occurring in a corpus precisely once (i.e., hapax legomena) and N_token is the total token frequency of words created by that morphological process in the corpus.

hapax legomenon (pl. legomena): a word form occurring only once in a corpus

Problems with this?

4. Productivity is the rate of addition of words using a morphological process over a period of time

How can you calculate this?

So though there are a number of measures, they're mostly indirect.

Beyond productivity

Because pretty much all morphological processes are only partly productive (if at all), sometimes there are competing members of paradigms. These can be studied and their contexts of use described.

• Within the same dialect
• slow vs. slowly, salesman vs salesperson
• Across dialects

• spoiled vs. spoilt, thought vs. thinked vs. thunk, envision vs. envisage

• Across time
• kine vs. cows, shit(ted) vs. shat

Also related to productivity are a variety of questions

• What is the base to which a morphological process applies?
• -ish, re-, un-, en-
• When do you use X-er and when more X?
• What morphological processes are optional, and what influences their use?
• -ly, subjunctive

Even pretty much unproductive stuff can have psychological reality, like phonaesthemes (Bergen 2004):

• Statistically frequent form-meaning parings
• gl- in words pertaining to light and vision: glisten, glitter, gleam, glow, glimmer, and many others;
• -ap in words about abrupt contact between surfaces: tap, slap, snap, clap, whap, bap, rap, etc.
• But non-compositional: -isten, -eam, and cl- don't contribute to the meanings
• And non really productive: there's no glap meaning to see contact between two surfaces.
• But you can see that they're part of the language through a corpus study.
• Freq_phon/Freq_meaning

the frequency of words that have the phonaesthemes in question over the total number of words with a similar meaning

• Freq_phon/Freq_form

the frequency of words with the phonaestheme over the total number of words with the same form.

• These can be compared to the expected values (the frequency of all words with the same form or meaning, compared to the total number of words in the corpus)

Or you can look at other statistics internal to words, such as the noun and verb business we saw last time.

References

Bauer , Laurie. 2001. Morphological productivity. New York and Cambridge: Cambridge University Press.

Bergen, Benjamin. 2004. The psychological reality of phonaesthemes. Language 80(2).