Retrieval of Lexical-Syntactic Features in Tip-of-the-Tongue States

1
Retrieval of Lexical-Syntactic Features in
Tip-of-the-Tongue States

• Michele Miozzo and Alfonso Caramazza
• Presented by Ping Yu

2
Tip of the tongue (TOT)

• Something is at the edge of your memory but you
  cannot produce it.
• But you feel it is right on the tip of your
  tongue
• It happens to most of us.
• It is a temporary inability to produce the
  word/phrase, but the word/phrase might be
  retrieved later.

3
William James (1842 -1910)a pioneering
psychologist and philosopher

• "It is a gap that is intensely active. A sort of
  wraith of the name is in it, beckoning us in a
  given direction, making us at moments tingle with
  the sense of our closeness and then letting us
  sink back without the longed-for term."

4
TOT meaning and form breakdown

• The word or name is in your memory (your mental
  lexicon).
• You cannot pronounce the whole word or phrase
• But you may be able to make some good guesses,
  e.g.
• Hmmm, the word starts with a b
• I know, it has three syllabus
• It ends with a d

5
So

• Two accounts of TOTs
• From aspect of phonology retrieval
• From aspect of lexical retrieval
• Miozzo and Caramazza (1997) discussed the lexical
  retrieval in TOT states. Specifically, they
  discussed retrieval of grammatical gender, the
  initial and final phonemes of Italian speakers in
  TOT states.

6
Two stages in speech production models
concept
speech
lemma
lexeme
Semantically syntactically specified
representation
Phonological representation
Less-studied area
7
Lemma and lexeme

• Lemma
• Put, e.g. John put a book on the table.
• Argument features
• Subcategorization features
• Tense

• Lexeme
• Number of morphemes
• Number of syllables
• Stress
• Number of phonemes

8
An Example
9
How the two stages are related?Nonoverlapping
serial-access model
lemma
lexeme
/k/, /au/
10
Bidirectional model Interaction-activation model
forward activation
lemma
lexeme
backward activation
/k/, /au/
11
Whatever model it is

• There is no agreement on the specific roles of
  and detailed access to the two stages.
• Though grammatical, semantic and phonological
  information are related in lexical access, there
  is no agreement on such fundamental issues as the
  overall number of representations, and how they
  are organized.
• It is agreed that grammatical information is
  specified at the level of lemmas. However, it is
  not agreed how the grammatical information is
  related to other representations and how it is
  retrieved.

12
The most clearly proposalbased on
serial-processing model
Syntactic node (grammatical category gender)
Morpho-syntactic properties
activate
Conceptual information
spread automatically and immediately
lemma an abstract lexical node
Lexeme Morphological and phonological properties
13
Some evidence

• Brain-damaged patients
• Some patients are able to retrieve nouns but not
  verbs, vice verse for others.

14
TOTs and the two stage models

• TOTs are often cited as evidence in support of
  the two-stage retrieval of lemmas and lexemes.
• TOTs demonstrate that the retrieval of meaning is
  independent of its form.
• TOTs reinforce feeling of knowing since many
  phonological features can be retrieved
  successfully.

15
Some facts and arguments

• Fact alternate words retrieved in TOTs are of
  the same grammatical class as the target word
  indicates that the targets grammatical category
  is recoverable.
• Morphosyntactic information associated with a
  specific lemma is available during TOT states
  when the corresponding lexeme is not available

16
But

• Those grammatical features such as tense, number,
  etc can be easily retrieved from context.
• During TOT experiments, subjects are usually
  asked to give definitions of infrequent words.
  The class of the retrieved word can be derived
  from definitions.
• So an ideal method is to demonstrate some
  syntactic features, that are not encoded in the
  meaning of the words, can be retrieved during TOT
  states.

17
So here is the paper

• Used as evidence something that has no meaning
  but has syntactic features.
• The assumption grammatical gender (in Italian)
  has no meaning.
• Goal can the grammatical gender be retrieved
  during TOT since it has no meaning derivable from
  lemmas?

18
In the literature

• The answer is yes.
• Vigliocco et al. (1997) addressed the TOT states
  in which Italian speakers know the gender of a
  word that they cannot produce in a naming task.
• Miozzo and Caramazza (1997) reached the same yes
  answer, but they
• tried to avoid the bias in TOT experiments
• reached some different conclusions

19
Grammatical gender

• Grammatical gender can be assigned in Italian,
  French, Spanish, German, and Dutch, etc.
• It is a syntactical category, not a semantic
  category. That is, it is independent of the
  semantics of its noun.
• Fiore (flower) is masculine in Italian but fleur
  (flower) is feminine in French.
• In Italian, tavolo (table) is masculine, and
  sedia (chair) is feminine.
• Although in Italian, word endings are indicators
  for gender types, e.g. a for feminine singular
  and o for masculine singular. The correlation is
  far from perfect. Many nouns whose gender is not
  determined from phonological structures.

20
Vigliocco et al (1997)

• They ask participants to guess the number of
  syllables and as many phonemes as possible in a
  naming task in TOT states.
• They found that the retrieval of gender in TOTs
  has nothing to do with the word endings since
  that participants can do equally well on guessing
  gender of those words whose gender do not
  correlate with the ending of the word.

21
Detailed findings of Vigliocco et al (1997)

• Italian speakers in a TOT state could
  successfully report the gender of the target word
  in a considerable proportion of times (84).
• The availability of gender information is not
  related to the availability of phonological
  features of the word, such as number of syllables
  and phoneme identity.
• This is consistent with the two-stage lexical
  models. That is, activation of lemmas precede the
  access of lexemes.
• Syntactic information can be retrieved prior to
  the retrieval of phonological information.

22
Critical

• Participants are searching for the word expected
  by the experimenter. A word other than that is
  considered noise. 41 of the total number of
  naming words were different from experimenters
  words.

23
Make-up method by Vigliocco et al(1997)

• They have used a post-TOT recognition test
  designed to assess the proportion of cases in
  which participants were seeking the wrong word.
• After TOT, they presented the target word to
  participants. Therefore, TOT states were divided
  into two groups
• positive TOTs, where the target word is the one
  participants pop in mind
• negative TOTs, where there is a mismatch.
• They found that grammatical gender was correctly
  retrieved far more frequently for positive TOTs
  (84) than for negative TOTs (53).

24
Bias in post-TOT recognition

• Participants were oriented towards the target
  words if they figured out the right phonological
  features.
• They were oriented towards a different word if
  they didnt figure out right phonological
  features.

25
Miozzo and Caramzza (1997)

• They tried to avoid the bias by establishing more
  firmly, with a different methodology of the same
  question that Vigliocco et al (1997) addressed.
• They tried to compare more directly the
  relationship between the gender and the
  phonological information retrieved in TOT states.

26
Goals of Miozzo and Caramzza (1997)

• How grammatical features are encoded
• How the retrieval of these features interacts
  with the retrieval of phonological forms.

27
Experiment 1

• To demonstrate that gender is not derived from
  the phonological characteristics of the noun in
  TOT states.
• -- by evaluating whether retrieval of gender
  information was correlated with retrieval of
  final vowel in TOT states.

28
Method

• 16 native Italian speakers (staff and students of
  the University of Padua)
• A total of 160 uncommon nouns (80 masculine and
  80 feminine) were selected for the naming task
• All target words are singular words ending with a
  vowel.
• For both masculine and feminine words, there were
  43 regular and 37 irregular words.

29
Procedure
A choice between gender A choice for the final
vowel Page 1
Picture-naming task Definition-naming task
15 s
TOT
Feel like knowing
Get the right answer
Dont know
A choice for the initial vowel or consonant
Page 2
Recovered TOT
30
Results

• 60.8 of the alternative names produced by
  participants are considered correct.
• Altogether there were 304 TOTs.
• Gender can be correctly selected despite the
  failure of retrieval relevant phonological
  features (final phonemes).
• In 71.1 of trials in TOTs, participants guessed
  the right gender.
• Performance in choosing gender is much better
  than in choose final phonemes (62.4).
• Gender regularity doesnt affect the gender
  choice.

31
Theoretical issues

• The retrieval of gender does not depend on the
  retrieval of final phonemes --- the part of a
  word that could potentially help in retrieving
  gender information. This is coincident with the
  two stages in lexical models.
• However, 76.6 of initial phonemes can be
  retrieved successfully. In the discrete two-stage
  lexical model, phonological information cannot be
  retrieved at the level of lemma.

32
Experiment 2

• Vigliocco et al. (1997) discarded all the
  non-target responses.
• In Miozzo and Caramazza (1997) only 6.3
  responses in TOT were non-taget nouns.
• Experiment 2 tried to further reduce the
  uncertainty in TOT experiments.
• The goal is to demonstrate that some features
  available in TOT states are not recalled by
  chance.

33
Method

• 32 native Italian speakers (staff and students of
  the University of Padua) participated. They
  didnt participate Experiment 1.
• The list of nouns was polished a little bit.
  Those words that were likely to have alternative
  non-target words were discarded. In this way,
  uncertainty was decreased.
• The same procedure was taken as Experiment 1
  except for dont know (DK) states, participants
  needed to choose the gender, the initial
  phonemes, the final phonemes for DK states as
  well.

34
Results
35
Analysis on bias

• Bias depends on whether the probability of
  correctly recognizing gender was affected by a
  targets gender or by a targets gender
  regularity.
• In the corpus of the 3000 most Italian common
  words, 63 are masculine.

36
Some comparisons between TOT and DK states

• In TOT responses, 72.4 masculine and 62.3
  feminine.
• Among corrected recognized gender, 69.3 were
  irregular nouns and 65.3 were regular nouns.

• In DK responses, 63.9 masculine and 36.4
  feminine bias consistent with the gender
  distribution in Italian
• Among corrected recognized gender, 60.7 were
  irregular nouns and 43.8 were regular nouns.
  (unclear bias source)

For final phoneme response, regular gender vowels
were more frequently selected. This bias occurred
in both TOT and DK responses. This bias occurred
in Experiment 1 as well.
37
Conclusion

• Miozzo and Caramazza (1997) replicated and
  extended the availability of grammatical gender
  in TOTs and some conclusion in Vigliocco et als
  (1997) study.
• In TOTs, despite the unavailability of the word
  form, the syntactic (and some phonological forms)
  features of the target words can be retrieved.
• Grammatical representation is independent of
  phonological representation. This is consistent
  with the with the two-stage lexical models, in
  which, semantic, syntactic and phonological
  features are independent.

38
But How about initial phonemes?

• Grammatical information can be retrieved better
  than initial phonemes since it is prior to the
  retrieval of phonological information.
• The results challenge the hypothesis of a strict
  dependence between the retrieval of grammatical
  and phonological information.

39
Also

• The retrieval of initial phonemes undermines the
  statement in some serial models, that is, the
  access to lemma leads to access to its syntactic
  features automatically and immediately.

40
My questions

• Does Italian gender really bear no meaning? Is it
  really a grammatical/syntactical category?
• Refer to Phillips and Boroditsky (2003)
• Can the retrieval of initial phonemes be
  explained by bidirectional model?

41
Selected references

• Some slides related to background knowledge are
  adapted from
• http//www.psychol.ucl.ac.uk/language/papers/pagan
  elli.pdf
• http//www.cogsci.rpi.edu/CSJarchive/Proceedings/2
  003/pdfs/180.pdf
• http//www.indiana.edu/rcapub/v17n1/24sb.html
• http//www.let.uu.nl/Frank.Wijnen/personal/neuroc
  og_of_lang/intro-lecture.pdf
• http//www.indiana.edu/ascpost/PowerPointpres/Bur
  ke_talk.PDF