LanguageWords and Rules

1
Language-Words and Rules

• The arbitrary sound-meaning pairing underlying
  words handled by associative memory.
• The discrete combinatorial system underlying
  grammar handled by symbol manipulating rules.
• These principles implicate distinct cognitive
  mechanisms is the bottom line of this theory

2
Contending Theories

• Generative phonology Invokes minor rules to
  generate irregular as well as regular forms
• Connectionism Invokes a pattern-associator
  memory to store and retrieve regular as well as
  irregular forms

3
The Problem-statement

• We ought to find a case in which words and
  rules express the same contents -- but they would
  still be psychologically, and ultimately
  neurologically, distinguishable.

4
Regular and Irregular Inflection

• Regular Inflection
• Vpast -- Vstem d
• Open-ended
• Predictable

• Irregular Inflection
• Form their past tense in idiosyncratic ways
• Only 180 irregular verbs in English
• Unpredictable Sink-sankslink-slunk
  (slank?)
• Think-thought (thunk?)blink-blinked
  (blought?)

5
Interaction of the Sub-systems

• Regular inflections are Hallmark of rule products
  while irregulars are of memorized words.
• If a word can provide its own past tense form
  memory, the regular rule is blocked.

6
How Irregular are irregulars?

• Patterns among irregular verbs
• Eg.1.Keep-sleep-dream-feel
• 2.Bear-wear-swear-tear
• So the irregular forms are not just a set of
  arbitrary exceptions, memorized individually by
  rote, and therefore cannot simply be attributed
  to a lexicon of stored items, as in the word-rule
  theory

7
Explaining Irregularity

• Generative Phonology
• Minor rules are used for irregulars.
• eg. ring-sing-sit vowel change i to a
• But doesnt apply to string, cling etc
• Also, Bring-brought
• Stick-stuck
• Spin-spun
• --The above exemplify that no rule can embrace
  all of these.

8
Contd.,

• ConnectionistsFeatures of one item are linked
  to features of other.
• Similar items, which share features, are partly
  superimposed in the memory representation,
  allowing the common patterns to reinforce each
  other.
• New items that are similar to learned items will
  activate the shared features allowing for
  generalization.

9
The Pattern Associator Model-Connectivism

• Input Nodes Phonological features of input stem
• Output Nodes Identical bank of output units
  representing the past tense form.
• Working
• 1.Every input node is connected to every
  output node.2.During training sessions strengths
  of various connections are raised and
  activated to varied extents.3.The past tense
  form is computed as the word that best fits the
  active output nodes.
• From now on this model replaces the
  connectionists point of view.

10
Words and Rules Theory

• Regular inflection is computed by a memory
  operation that does not need access to the
  contents of memory unlike irregular inflection
  which come from memory.
• The evidence In circumstances in which memorized
  forms are not accessed, for one reason or
  another, irregular inflection should suffer and
  not regular.

11
1.Weak Memory entry (rare word)

• Low frequency irregulars are easily lost old
  English had double the irregulars than present.
  cleave-clove, crow-crew, abide-abode, chide-chid,
  and geld-gelt.
• smite-smote, slay-slew, bid-bade, spell-spelt,
  and tread-trod (in American English) are
  predicted to be lostlook out!
• In some cases a form is familiar enough to block
  the regular version, but not quite familiar
  enough to sound natural, example, in a cliché,
  idiom etc. which is used in a characteristic
  tense.

12
The million word-corpus

• 1. be 39175/million
• 2. have 12458
• 3. do 4367
• 4. say 2765
• 5. make 2312
• 6. go 1844
• 7. take 1575
• 8. come 1561
• 9. see 1513
• 10. get 1486

13
The bottom ten

• 3791. abate 1/million
• 3791. abbreviate 1
• 3791. abhor 1
• 3791. ablate 1
• 3791. abridge 1
• 3791. abrogate 1
• 3791. acclimatize 1
• 3791. acculturate 1
• 3791. admix 1
• 3791. adsorb

14
Examples

• 1.You will excuse me if I forgo the pleasure of
  reading your paper before its published.
• ?Last night I forwent the pleasure of grading
  student papers.
• 2.I dont know how she can bear that guy. ?I
  dont know how she bore that guy.
• .

15
Contd.

• This unfamiliarity doesn't happen when regulars
  are used in infinitival forms
• 1.We cant afford it.
• I dont know how he afforded it.
• 2.She doesnt suffer fools gladly.
• None of them ever suffered fools gladly.
• That means, an irregular verb and past tense
  both are to be individually familiarized where as
  for regulars one of them would suffice.

16
2. Difficult-to-analogize (unusual-sounding)
verbs

• Kinds of unusual verbs
• 1. Prototypical verbs like plip sound like many
  existing English regular verbs.
• 2.Intermediate Verbs like smaig do not rhyme with
  any existing English verb root.
• 3. Unusual verbs like ploamph are phonologically
  illicit in English and hence are very dissimilar
  to existing verbs.

17
Contrastive responses Human an Model

• 1.plipped was the answer, without trouble.
• 2.Humans gave verbed for all strange verbs of 2
  and 3 categories.

• Gave plipped, but had a little trouble.
• 2. It created random combinations
  likesmairf-sprurice, trilb-treelilt,
  smeej-leefloag, and frilg-freezled.

18
Interpretation of failure

• Pattern associator models, do not have the
  mechanism of a variable, such as "Verb," that can
  stand for an entire class regardless of its
  content and that can thereby copy over the
  phonological material of a stem so that it can be
  systematically modified to yield a past tense
  form.
• They create a chimerical output by picking up
  syllables and applying strengthened connections.
• So, memory reinforcements cannot handle the new
  situation like humans who just add ed to every
  strange verb. It requires a separate
  symbol-manipulating system.

19
3.Irregular form is trapped in memory

• All my daughters friends are low-lifes
  (low-lives).
• Im sick of dealing with all the Mickey Mouses in
  this administration (Mickie Mice).
• The above examples show that sound alone cannot
  be the input to inflection.
• Semantics triggered this change?

20
Contd.,

• overate/overeated
• Workmen/workmans
• Gods children/childs
• Caught cold/catched
• So semantics doesnt guide this systematic
  regularisation. The explaination lies in the
  headless ness or righthand-rule.

21
Contd.,

• In case of head-less words, pipeline of
  information is broken because of
  non-connectivity.
• Over-eat is just another kind of eating but
  low-life is a person and not a kind of life.
• What about Mickey-mouse? The connection is
  between Proper noun Mickey mouse and general noun
  mouse.

22
6,7,8.Onomatopoeia, Quotations and Foreign
borrowing

• The engine pinged/panged
• While checking for sexist writing, I found three
  "man"s/"men" on page 1.
• succumbed/succame derided/derode
  chiefs/chievesgulfs/gulves (all borrowed from
  French or Latin).

23
Limitation on Regular Plurals

• Rat-infested vs Mice-infested
• Although rats and mice are semantically
  equivalent, rats-eater isnt a correct usage.
• teethmarks versus clawsmarks, men-bashing versus
  guys-bashing, and purple-people-eater versus
  purple-babies-eater

24
Explaining it

• Route in the mental architecture
• Memorized -- Complex -- Regular
• Roots Word Inflection
• (including Formation
• irregulars)
• The word mice, stored as a root in the first
  component, is available as an input to the
  compounding process in the second component,
  where it is joined to infested to yield
  mice-infested. In contrast, rats is not stored as
  a memorized root in the first component it is
  formed from rat by an inflectional rule in the
  third component, too late to be inputted to the
  compounding rule in the second

25
The Gordons experiment

• Peter Gordon (1985) showed that 3-5-year-old
  children are sensitive to this principle. He
  asked them questions such as, "Here is a monster
  who likes to eat X. What would you call him?"
  First he trained them on mass nouns such as mud,
  which dont take a plural, to introduce them to
  the compound construction, in this case
  mud-eater, without biasing their subsequent
  answers. Then he tested them by asking what they
  would call a monster who likes to eat rats. The
  children virtually always said rat-eater, not
  rats-eater. In contrast, they frequently called a
  monster who likes to eat mice a mice-eater -- and
  those children who occasionally used the
  overregularized plural mouses in other contexts
  never used it in a plural such as mouses-eater.

26
Inference

• That means one never compounds a regular
  inflection with a complex word and this need not
  be taught.
• Except for tooth brush none of these forms were
  common in parental vocabulary of the experimented
  children.
• Therefore childrens sensitivity to the
  teethmarks/clawsmarks distinction is likey to be
  a product of the innate architecture of their
  language system.

27
Disrupted memory-word retrieval

• "Everyday I like to verb. Yesterday I
• ________.
• This was the sentence and was experimented with
  novel verbs of three kinds I specified earlier.
• -Tested on two kinds of patients
• -Anomic aphasia
• -Agrammatic aphasia

28
Predictions

• Patients who are more impaired on vocabulary
  retrieval than on grammatical combination should
• (1) find irregular forms harder to produce than
  regular ones,
• (2) should occasionally produce
• overregularized forms such as swimmed (for the
  same reason that children do), and
• (3) should have little trouble producing past
  tense forms for novel verbs such as plammed.

29
Prediction

• Patients who are more impaired on grammatical
  combination than on vocabulary retrieval should
• (1) find regular forms to be harder to produce
  than irregular ones,
• (2) should rarely produce over regularized forms,
  and
• (3) should have grave difficulty producing past
  tense forms for novel verbs.

30
Results-Anomic aphosia

• As predicted, found irregular verbs harder to
  inflect than regular verbs (60 vs 89)
• made frequent overregularization errors (25 of
  the opportunities), and
• were fairly good with novel verbs (84)

31
Results-Agrammatic aphasia

• found regular verbs harder to inflect than
  irregular verbs (20 versus 69)
• made no over regularization errors, and
• were poor at inflecting novel verbs (5).
• Similar results were found by many researchers.

32
Cross-Linguistic Validation

• In Hindi, inflection for plurals is given based
  on the nucleus of the last syllable
• a-e/AyEn(laDkA,bAlikA)
• I-iyAn(titli)
• U-nil(bandhU)
• Consonant ending-en/-(imArat/patang)
• bandar, patang, ciDiyA etc. are mostly left
  un-inflected especially consonant endings.
• English words sometimes retain their plurals
• Maam ne bahut examples dI thI
• rOz kitni classes hOtI hain terI?

33
Contd.,

• There is one case in which semantics of the word
  did influence the inflection
• faTAkA-faTAkE
• imArat-imAratEn
• But, bAlikA-bAlikAyEn although a ending is
  inflected differently.
• For novel words I tried(kAdal,paDkA,miDsI,balpam,c
  EpagruDDU), most of the people either decided it
  based on the vowel or left it un-inflected.

34
Contd.,

• Pattern-associaton would be challenging if we
  introduce Hindi to them, because there is no
  reinforcement in many cases. But that doesnt
  mean there is no notion of plural in that case,
  because in the oblique case all those attain
  inflection E/iyOn/On/uon.
• Here reinforcement has taken place in mind of a
  cognate but not machineit cannot reinforce
  nothing
• Moreover, it would be more fruitful to experiment
  a memory-loss impaired person on un-modified
  plurals in Hindi as there is no burden on the
  memory at all.

35
Contd.,

• In case of German the regular suffix for
  past-tense -t is used only 45 times as against
  85 times in English. That is, irregular forms
  are the majority.
• But still, Germans add t suffix to novel verbs,
  onomatopoeia and borrowed verbs. notions of
  Frequencies and reinforcements of the
  connectionists is invalidated here. Psychological
  notion of default suffix is there.
• Compounding also, has similar effect across
  languages
• English rat-eater (rats-eater)
• German auto- fresser(autos-fressercar-eater)
• Hindi bhEnDi kI sabjI(bhEnDiyOn kI sabjIladys
  finger curry)

36
Conclusion

• It is now established that regular forms are
  default operations applying whenever memory
  retrieval fails to provide an inflected form.
  Regular inflection applies freely in any
  circumstance in which memory fails because
  regular inflection is computed by a mental
  operation that does not need access to contents
  of memory, namely, a symbol-processing rule.
• The evidence, then, supports the hypothesis that
  the design of human language comprises two mental
  mechanisms memory, for the arbitrary sign
  underlying words, and symbolic computation, for
  the infinite use of finite media underlying
  grammar.

37
Questions???

• Words and rules theory predicts that a grammarian
  of the great classical language of Sanskrit
  (which has no irregularities of any form in it)
  would continue to be proficient even if we blast
  his entire left posterior perisylvian regions!!!