Inf1 CogSci 2012: Lecture 8: The Broken Telephone

1. Summary of English verb inflection

English inflection is very simple.

Only suffixes; no prefixes (or even infixes!)
At most one inflection per verb, i.e. *opensed, *opensing

Therefore, we can posit just one rule capturing all English verb inflection:

A verb is composed of a stem and an optional suffix

The stems and suffixes themselves are stored in, and retrieved from, the mental lexicon.

2. Recap: Words and rules

Human language appears to involve two different kinds of "mental tissue":

a lexicon of words - arbitrary, memorised
a grammar of rules - productive, abstract, compositional, recursive

Evidence for this dichotomy comes from the contrast between regular and irregular past tense formation in English:

regular - "talked" is formed by a rule
irregular - "sang" is stored as a word

These two methods of past tense formation work together to yield a system which is:

expressive - every verb gets a past tense form
efficient - the most common past tense forms are stored as words and can be retrieved more quickly

3. Recap: Dissection by linguistics

Language is modular:

lexicon/morphology/syntax/phonology/semantics diagram

Reproduced under fair use

There are (at least) three kinds of rule:

morphological - combine stems and affixes into words
syntactic - combine words into phrases and sentences
phonological - rewrite sequences of sounds to make them easier to pronounce

4. Irregular verbs and broken telephones

All irregular past tense verb forms used to be regular!

they were formed by productive rules
e.g. "sing"-"sang", "ring"-"rang"

However: the process of language acquisition by children has slightly imperfect fidelity:

your language is slightly different from your parents' language, a little bit more different from your grandparents', etc. etc.
hence the "broken telephone" metaphor (a.k.a. Chinese Whispers)

Hypothesis:For any irregular past tense form in Modern English, some past generation of English speakers must have failed to grasp the relevant rule

they just memorised the past tense forms as separate words in their mental lexicon
they then passed this new version of the language on to their own children

Once a past tense form has been stored as a word, it can subsequently pick up more and more idiosyncracies

and hence become more and more irregular down the generations!

5. Example from plural inflection

Why is the plural form of "foot" not "foots" but rather "feet"?

In Old English, this was a regular plural, formed using the suffix [i]

[fot] + [i] = [foti]

Like all Germanic languages, Old English had a phonological rule called Umlaut:

Change a back vowel into a front vowel, when the next syllable contains a front high vowel (cf. vowel harmony)
Courtesy of AugPi
e.g. [foti] => [feti]
this means less work for the tongue
Umlaut still exists in Modern German, e.g. Kuh - Kühe

By Middle English, people had started to drop unstressed final syllables consisting of a solitary vowel:

e.g. [feti] ==> [fet]
hence, the plural of "foot" became irregular

I use ==> rather than => to denote diachronic change

Finally, the Great Vowel Shift changed all [o]'s into [u]'s and all [e]'s into [i]'s

[fot] ==> [fut] (i.e. "foot")
[fet] ==> [fit] (i.e. "feet")

6. Vowel shortening in irregular verb forms

Why is the past tense form of the verb "keep" not "keeped" but rather "kept"?

Similar examples:

weep-wept, creep-crept, sleep-slept, leap-lept, sweep-swept
feel-felt, deal-dealt, kneel-knelt, dream-dreamt, leave-left
bleed-bled, breed-bred, feed-fed, lead-led, plead-pled, read-read, speed-sped, meet-met
hide-hid, slide-slid, bite-bit, light-lit
lose-lost, shoot-shot

Again, these verbs used to all be regular.

During the Middle English period, many stems containing long vowels were shortened when suffixes were added:

child-children
deep-depth
five-fifth
wise-wisdom

This also applied to many hitherto regular past tense verb forms, rendering them eventually irregular:

keeped [kipəd] ==> [kɛpəd] ==> kept
losed ==> lost

7. Strong verbs

All the irregular verbs discussed so far started out as regular verbs in Old English

they all have some reflex of the past tense suffix [d]

The "strong" verbs are different:

they lack any sign of a past tense suffix
- e.g. bear-bore, sink-sank, rise-rose
they were already irregular in Old English

But they were regular verbs in Proto-Indo-European!

8. Proto-Indo-European

PIE was probably spoken in southern Russia (or eastern Turkey?) over 6000 years ago.

The earliest known ancestral language of English

and German, Dutch, Danish, Swedish and Icelandic
and French, Spanish, Irish, Welsh, Russian, Polish and Greek
and Hindi, Urdu and Farsi

PIE has been reconstructed by historical linguists by means of the "comparative method"

involving painstaking comparisons of the vocabularies of all these languages.

9. Conclusion

All the verbs which are currently irregular in English used to be regular:

many had regular past tense forms in Old English (formed by [d] suffixation), but became irregular during the Middle English period (suppletion, syncope, apocope, vowel shortening)
the strong verbs had regular past tense forms in PIE (formed by ablaut), but had already become irregular in Old English.

This explains why there appear to be patterns in irregular past tense inflections.

Irregular past tense forms are stored in the lexicon as words, independent of the verb's base form.

The longer a verb has been irregular in English, the more likely its past tense form will have picked up lexical idiosyncracies.

i.e. the more irregular it will have gotten.

10. Recap: Words, rules and irregular verbs

We now have a very simple theory of English past tense verb inflection:

regular forms are the result of applying a productive rule
irregular forms are stored in the lexicon as independent, memorised words

This theory has many advantages:

it explains why new verbs in English get regular inflection (e.g. mosh-moshed, snarf-snarfed)
it explains why children generally assume that verbs they haven't encountered before get regular inflection (e.g. rick-ricked, bing-binged)

However, the simple theory also has important deficiencies.

11. Irregular inflection is semi-systematic

We have seen that English past tense inflection is shot through with patterns:

blow-blew, grow-grew, know-knew, ...
bind-bound, find-found, grind-ground, ...
drink-drank, shrink-shrank, sink-sank, ...
bear-bore, wear-wore, swear-swore, ...

i.e. Suppletion (e.g. go-went) is the exception rather than the rule.

We have seen that these patterns are the fossils of rules that lived in the minds of Old English or PIE speakers, but are now extinct.

But: there is also evidence to suggest that these patterns are represented, in some way, in the minds of modern-day English speakers.

12. Diachronic evidence

Since Old English, some verbs have switched from being regular to being irregular:

e.g. ring-ringed ==> ring-rang
i.e. "ring" was attracted to the class of strong verbs by analogy with verbs like "sing-sang"
other examples: dig-dug, stick-stuck, wear-wore

Some verbs that entered the language after the time of Old English were seduced into strong verb classes, rather than receiving regular inflection:

e.g. fling-flung, sling-slung

Some verbs switched from one strong verb class to another:

e.g. draw-drough ==> draw-drew

Some verbs switched from being regular to being "weakly" irregular:

light-lighted ==> light-lit
creep-creeped ==> creep-crept

Moreover, some of these developments occurred relatively recently, i.e. in the 19th century:

kneel-kneeled ==> kneel-knelt
catch-catched ==> catch-caught
quit-quitted ==> quit-quit

13. Evidence from children

Young children occasionally over-generalise irregular past tense inflection:

"They swang into a rollercoaster" (i.e. swung)
"I shuck his coat" (i.e. shook)
"Doggie bat me" (i.e. bit)

In experimental recordings, 8 out of 9 children make at least one error of this kind.

Children make this kind of error well into their school-age years.

Children have never heard adults using past tense forms like "swang" or "shuck"

they must be constructing these forms creatively, by analogy with other verbs they already know.

14. The moral

Irregular verbs cannot simply be memorised by rote

since they show evidence of patterning
even in the minds of modern-day English speakers!

Thus, the distinction between regular and irregular inflection, and hence between words and rules, is not so clear anymore.

There are two obvious ways of resolving this paradox:

irregular past tense forms are also generated by rules
there are no rules, only a general associative mechanism for recognising patterns.

The first approach is associated with Chomsky and Halle's generative phonology.

The second with Rumelhart and McClelland's parallel distributed processing (or "connectionism").

15. Rationalism versus empiricism

The distinction between generative phonology and PDP is a recent development in a 500 year old philosophical debate about the nature of the human mind.

Rationalism - intelligence arises from the manipulation of symbols by rules

associated with Leibniz and Descartes
and more recently with Noam Chomsky
the human mind has lots of innate structure
knowledge comes from logical deduction (i.e. "calculation").

Empiricism - intelligence arises from the mind connecting together things that were experienced together or that look alike

associated with John Locke and David Hume
more recently with behaviourism (Ivan Pavlov, B.F. Skinner - stimulus/response)
and even more recently with connectionism
the human mind starts out as a "blank slate"
knowledge comes by generalising from observations.

Can the study of regular and irregular past tense inflection in English resolve one of the most important controversies in philosophy?