Interactions between semantics and phonology in the relatedness decision task

1
Interactions between semantics and phonology in
the relatedness decision task

• Anna Woollams
• Department of Psychology, University of
  Wollongong
• Sachiko Kinoshita
• Macquarie Centre for Cognitive Science, Macquarie
  University

2
Phonological Mediation

• To what extent is phonology is involved in
  accessing word meaning during silent reading in
  skilled adults?
• Access to meaning from print occurs exclusively
  via phonology (Van Orden, 1987).
• Do phonological factors affect performance in
  semantic decision tasks?

3
Homophones

• Jared Seidenberg (1991, E2) demonstrated that
  the homophone effect is confined to low frequency
  foils with low frequency exemplars
• SUN (son) SIN NIGHT (knight) KNIFE
• ROWS (rose) ROBE FLEE (flea) gt FLEX

4
Phonological Mediation

• The homophone effect could arise from whole-word
  or sub-word phonological activation of meaning.
• Dual-route models (Coltheart et al., 2001)
  incorporate two separate mechanisms for
  computation of whole-word and sub-word phonology,
  hence they are independent effects.
• Single-route models (Plaut et al., 1996)
  incorporate one common mechanism for computation
  of whole-word and sub-word phonology, and
  therefore these effects should co-occur.

5
Pseudohomophones

• Jared Seidenberg (1991, E2) demonstrated that
  the pseudohomophone effect is confined to foils
  with low frequency exemplars
• CHEEF (chief) CHEIL WIRM (worm) gt WURN

6
False homophones

• Lesch Pollatsek (1998, E1) obtained a
  significant false homophone effect in relatedness
  decision.
• Higher frequency true associates were associated
  with longer reaction times to false homophones.

7
Sub-Word phonology

• Pseudohomophone and false homophone effects
  reflect sub-word phonological activation of an
  inappropriate semantic representation.
• Does sub-word phonology affect meaning access if
  it does not activate an inappropriate semantic
  representation?
• Are spelling-sound typicality effects seen in
  semantic decision tasks?

8
Spelling-Sound Typicality

• Although false homophones were less consistent
  than spelling controls, Lesch Pollatsek (1998,
  E3) did not obtain a significant consistency
  effect for unrelated trials.
• However, many of the false homophones had only a
  few low frequency enemies.

9
Spelling-Sound Typicality
Definable meaning or given name?

• Taft van Graan (1998) did not observe a
  typicality effect in semantic categorization for
  words that had produced a significant typicality
  effect in naming.
• As word meaning modulates typicality effects in
  naming, perhaps typicality functions to modulate
  effects of word meaning in semantic decision
  tasks?

10
Experiment 1

• Factorial manipulation of imageability and
  body-rime consistency in a relatedness decision
  task (N80).
• Typicality was defined as Token FFCR (0.98 vs
  0.24).
• Targets were of medium frequency (37 per
  million).
• Context words were synonyms or associates of the
  monosyllabic targets, matched for length and
  frequency.
• Relatedness ratings, written frequency and number
  of phonemes were used as covariates in item
  analyses.

11
E1 Related Trials
12
E1 Related Results
13
E1 Summary

• No evidence of reliable consistency effects in
  related (or unrelated) decisions, as per previous
  studies.
• The imageability effect was only reliable for
  consistent words in both reaction time and error
  rate.
• Correct sub-word phonological activation
  facilitates semantic access and/or incorrect
  sub-word phonological activation inhibits
  semantic access.

14
Experiment 2

• Gottlob et al. (1999, E3) found an ambiguity
  disadvantage in relatedness decision for trials
  with context words related to both dominant and
  subordinate meanings.
• These ambiguity disadvantages could have resulted
  from confounds in meaning frequency that arose as
  a result of matching unambiguous words on written
  frequency.
• Will any ambiguity disadvantages observed with
  meaning frequency matched unambiguous words be
  modulated by consistency?

15
Experiment 2

• Manipulated ambiguity and body-rime consistency
  in a relatedness decision task (N120).
• Typicality was defined as Token FFCR (0.69 vs
  0.49).
• Ambiguous targets were of medium frequency (38
  per million) with an average dominance ratio of
  0.76.
• Context words were synonyms or associates of the
  monosyllabic targets, matched for length and
  frequency.
• Relatedness ratings, written frequency and number
  of phonemes were used as covariates in item
  analyses.

16
E2 Related Dominant Trials
17
E2 Related Dominant Results
18
E2 Related Subordinate Trials
19
E2 Related Subordinate Results
20
E2 Summary

• No evidence of reliable consistency effects in
  related (or unrelated) decisions, as per previous
  studies.
• The ambiguity disadvantage for subordinate trials
  was only reliable for consistent words in both
  reaction time and error rate.
• Correct sub-word phonological activation
  facilitates semantic access and/or incorrect
  sub-word phonological activation inhibits
  semantic access.

21
Conclusions

• Spelling-sound typicality determined the
  magnitude of imageability and ambiguity effects
  in relatedness decision.
• Sub-word phonological activation does affect
  meaning access even when it does not correspond
  to an inappropriate semantic representation.
• Effects of sub-word phonological activation are
  clearly weaker when they do not correspond to an
  inappropriate semantic representation.

22
Phonological Mediation

• Do phonological factors affect performance in
  semantic decision tasks?
• Strongest if activates an inappropriate semantic
  representation.
• Access to word meaning from print occurs
  exclusively via phonology (Van Orden, 1987).
• Direct access from orthography occurs for high
  frequency words.
• To what extent is phonology is involved in
  accessing word meaning during silent reading in
  skilled adults?