Date post: | 21-Jan-2016 |
Category: |
Documents |
Upload: | anis-fletcher |
View: | 216 times |
Download: | 0 times |
Is the pan without eggs the same as the empty pan?
How types of modifiers affect choice of referring expression and memory
Danielle Moed
MA Linguistics, University of Toronto
Psycholinguistics Group – Nov 7, 2014
Danielle Moed - University of Toronto - November 7, 2014 2
Overview
• Introduction– Purpose– Lexical versus Contextual Modifiers– Psycholinguistic Research– Representation of Alternatives– Memory Consequences for Inferring Alternatives– Present Study
• Method
• Results
• Discussion
Danielle Moed - University of Toronto - November 7, 2014 3
INTRODUCTION
4
What would you call the highlighted bag?
• Referring expressions differ in what information they provide about an entity and indirectly about other entities in the context
Danielle Moed - University of Toronto - November 7, 2014
Danielle Moed - University of Toronto - November 7, 2014 5
What about these bags?
• Referring expressions do not only reflect the properties of the entities they are used to refer to, but can potentially also be used to infer the properties of other entities
Danielle Moed - University of Toronto - November 7, 2014 6
PURPOSE
Danielle Moed - University of Toronto - November 7, 2014 7
Purpose
• Investigate how and to what extent different referring expressions affect the representation of the properties of unmentioned entities by examining how the production of one referring expression or another affects subsequent memory of these unmentioned entities
Danielle Moed - University of Toronto - November 7, 2014 8
LEXICAL VERSUS CONTEXTUAL MODIFIERS
Danielle Moed - University of Toronto - November 7, 2014 9
Maxim of Quantity
• A speaker should only be as informative as required for the current purposes of the conversation
Danielle Moed - University of Toronto - November 7, 2014 10
Lexical Modifiers
• Adjectives open, empty, …
• Lexical in the sense that the existence of alternatives is part of the lexical meaning of these adjectives
• Gradable Adjectives
Danielle Moed - University of Toronto - November 7, 2014 11
Gradable Adjectives
• Denote properties that are relative to a comparison class (Kennedy, 2007; Kennedy & McNally, 2005).
The wine glass is full.
Danielle Moed - University of Toronto - November 7, 2014 12
But wait!
• Important to note that not all adjectives are lexical modifiers
• Alternatives are not part of the lexical meaning of all adjectives
• Esp. those adjectives that denote colour, shape, pattern and material – These can be considered to be contextual
Danielle Moed - University of Toronto - November 7, 2014 13
Non-lexical Adjectives
• Could be said to be contextual
Danielle Moed - University of Toronto - November 7, 2014 14
Contextual Modifiers
• PPs without asparagus, …
• Contextual in the sense that the lexical meaning of the modifier does not encode the alternative, but rather an alternative can be inferred from the context
Danielle Moed - University of Toronto - November 7, 2014 15
PSYCHOLINGUISTIC RESEARCH
Danielle Moed - University of Toronto - November 7, 2014 16
Contrastive Interpretation in Context Despite Lexical Differences
• Sedivy, Tanenhaus, Chambers and Carlson (1999)– Visual-world eye-tracking paradigm to determine how and when
listeners use a contextually available contrast set when interpreting non-contrastive adjectives (i.e., colour/size)
• Result: – 1B: Eye-movement latencies and the likelihood of looks to the non-
target objects indicated that listeners initially expect a contrastive interpretation for non-contrastive adjectives if there was another object in the display from the same nominal category
– 2: Results of this experiment revealed that listeners interpreted the scalar adjective contrastively, anticipating the referent to be apart of a contrast set.
Danielle Moed - University of Toronto - November 7, 2014 17
But…
• Heller and Chambers (2014) found differences between adjectives that encode a comparison class and those that do not
• Examined reference to objects that contrasted along two dimensions (i.e., colour and size), examining how an earlier produced referring expression for an object affects later reference to another object
• Results showed that that colour adjectives (e.g., blue) were less sensitive to the comparison class compared to size adjectives (e.g., large).
Danielle Moed - University of Toronto - November 7, 2014 18
Sedivy et al. VS. Heller & Chambers
• Sedivy et al., (1999) who put colour and size adjectives on unequal footing and found that listeners expected alternatives for colour and size adjectives
• Heller and Chambers (2014), who put colour and size adjective on equal footing in their experiment, found that colour adjectives were less likely than size adjectives to introduce alternatives
Danielle Moed - University of Toronto - November 7, 2014 19
What does this suggest?
• These results suggest that even when the visual context provides explicit alternatives, the inherent semantic contrastiveness supersedes the contextual influence.
Danielle Moed - University of Toronto - November 7, 2014 20
REPRESENTATION OF ALTERNATIVES
Danielle Moed - University of Toronto - November 7, 2014 21
Inferring Alternatives
• Computing alternatives involves combining linguistic information with information from context and world knowledge
• Normally alternatives are not explicitly given, must be inferred
Danielle Moed - University of Toronto - November 7, 2014 22
Explicit Alternatives
• But Cohen (1999) notes that alternatives can be explicit!
• Raises interesting question about the similarities and differences between inferred and explicit alternatives
Danielle Moed - University of Toronto - November 7, 2014 23
MEMORY CONSEQUENCES FOR INFERRING ALTERNATIVES
Danielle Moed - University of Toronto - November 7, 2014 24
Memory & Alternatives
• Testing for memory can reveal what has been computed, specifically showing which alternatives are available
• Fraundorf, Watson & Benjamin (2010)
Danielle Moed - University of Toronto - November 7, 2014 25
Contrastive Stress & Memory
A. Both the British and the French biologists had been searching Malaysia and Indonesia for the endangered monkeys.
B. Finally, the (British/French) spotted one of the monkeys in (Malaysia/Indonesia) and planted a radio tag on it.
Danielle Moed - University of Toronto - November 7, 2014 26
Contrastive Stress & Memory
• Results from a recognition memory tests indicated that contrastive stress increased the likelihood that listeners would remember the correct statements (i.e., the likelihood of hits), as well as the likelihood of correct rejections.
• Suggests that alternatives are inferred and represented, leading to better memory for those alternatives
Danielle Moed - University of Toronto - November 7, 2014 27
PRESENT STUDY
Danielle Moed - University of Toronto - November 7, 2014 28
The three kinds of modifier types
Danielle Moed - University of Toronto - November 7, 2014 29
METHOD
Danielle Moed - University of Toronto - November 7, 2014 30
Participants
• Present data from 30 English speaking participants from the University of Toronto Community
• 35 participants tested, but 5 were excluded from analysis
Danielle Moed - University of Toronto - November 7, 2014 31
Materials and Procedure
• Study contained 3 tasks:1. Referential Communication Task
2. Distractor Task
3. Memory Task
Danielle Moed - University of Toronto - November 7, 2014 32
REFERENTIAL COMMUNICATION TASK
Danielle Moed - University of Toronto - November 7, 2014 33
Experimental Materials
• Experimental Trials (18)– 6 in each of the 3 conditions:
• i.e., 6 contextual, 6 lexical & 6 both
– Each display contained 4 images• 2 formed pair (referent & alternative)• 2 distractor images
Danielle Moed - University of Toronto - November 7, 2014 34
Experimental Display
Danielle Moed - University of Toronto - November 7, 2014 35
Lists• 3 lists (List A, B, C)
– Eliminate overlap of experimental items
• Example: “Bag”List A: Lexical List B: Contextual List C: Both
Referent Alternative Referent Alternative Referent Alternative
Danielle Moed - University of Toronto - November 7, 2014 36
Filler Materials
• Filler Trails (82)– Similar to experimental displays
– Each display contained 4 images• 2 formed pair• 2 distractor images
– 26 trials required modification• Referent was a member of pair
– 54 trials did not require modification• Referent was a distractor image
Danielle Moed - University of Toronto - November 7, 2014 37
Filler Display: Modification
Danielle Moed - University of Toronto - November 7, 2014 38
Filler Display: No Modification
39
Procedure• Participant played a game with the experimenter to get
the experimenter to click on the picture
Participant
Experimenter (Addressee)Click on
the empty bag
Danielle Moed - University of Toronto - November 7, 2014
Danielle Moed - University of Toronto - November 7, 2014 40
DISTRACTOR TASK
Danielle Moed - University of Toronto - November 7, 2014 41
Distractor (Math) Task
• Materials:– 12 math sheets ( 4: +, 4: - & 4: +/-)
• Procedure: – Complete as many +/- equations as possible
in 10 minutes
Danielle Moed - University of Toronto - November 7, 2014 42
MEMORY TASK
Danielle Moed - University of Toronto - November 7, 2014 43
Experimental Materials
• 1 image per display (from relevant list)
• 2 displays for each experimental trial – 1 “seen” image – 1 foil image
• Example (Both condition)
Danielle Moed - University of Toronto - November 7, 2014 44
Seen Image Example
Danielle Moed - University of Toronto - November 7, 2014 45
Foil Image Example
Danielle Moed - University of Toronto - November 7, 2014 46
Lists
• Related to Referential Communication Task (i.e., List A, B, C)
• Arranged in 2 blocks (e.g., List A1, A2…)
Danielle Moed - University of Toronto - November 7, 2014 47
Filler Materials
• 1 image per display
• 2 displays for each experimental trial – 1 “seen” image – 1 foil image
• R Trials: “seen” image = referent• D Trials: “seen” image = distractor image
Danielle Moed - University of Toronto - November 7, 2014 48
Procedure
• Participants click “I saw this” or “I did not see this”
I saw this I did not see this
Danielle Moed - University of Toronto - November 7, 2014 49
RESULTS
Danielle Moed - University of Toronto - November 7, 2014 50
REFERENTIAL COMMUNICATION TASK
Danielle Moed - University of Toronto - November 7, 2014 51
Predictions
• Lexical Condition: Preference for producing lexical modifiers (e.g., open bag)
• Contextual Condition: Preference for producing contextual modifiers (e.g., bag with corn)
• Both Condition: No preference for producing lexical or contextual modifiers (e.g., empty bag/bag without asparagus)
Danielle Moed - University of Toronto - November 7, 2014 52
Syntactic Form Analysis• Bare Nouns: whether the referring expression only included a noun in addition to
the determiner (e.g., the flower) or the noun was a compound noun (e.g., the walking sign). 1 iff the noun is bare and 0 iff the noun is modified.
• Adjectival Phrase (AP): whether the referring expression contained a prenominal
adjectival modifier (e.g., the yellow flower). 1 iff the modifier was an adjective and 0 iff the modifier was not an adjective. Note that if the adjective appeared in a post-nominal relative clause (e.g. the vase that’s broken), it was coded as 0 here.
• Prepositional Phrase (PP): whether the referring expression contained a modifier
that is a prepositional phrase (e.g., the vase without a flower). 1 iff the modifier was a PP and 0 iff the modifier was not a PP.
• Relative Clause (RC): whether the referring expression contained a relative clause
(e.g., the vase that’s broken). Note that the relative clause could contain an adjective (e.g., the bag that was open). 1 iff the modifier was a relative clause and 0 iff the modifier was not a relative clause.
Danielle Moed - University of Toronto - November 7, 2014 53
Mean likelihood of the different modifier forms by condition
Danielle Moed - University of Toronto - November 7, 2014 54
Semantic Form Analysis
• Lexical Modification: 1 iff the modifier in the referring expressions was such where alternatives are part of the lexical meaning of that modifier. For example, gradable adjectives like open or empty whose interpretation depends on a comparison class.
• Contextual Modification: 1 iff the modifier in the referring expression was such that the information in the modifier allowed discriminating the images in the context, but its lexical meaning did not imply an alternative. This includes cases where the modifier explicitly mentions a property of the referent (e.g., bread with butter), or where it implies a property of the alternative (e.g., bread without peanut butter). In both cases, one can only infer that the alternative does not have the property of the referent.
Danielle Moed - University of Toronto - November 7, 2014 55
Mean likelihood of lexical and contextual modifiers
Danielle Moed - University of Toronto - November 7, 2014 56
Danielle Moed - University of Toronto - November 7, 2014 57
MEMORY TASK
Danielle Moed - University of Toronto - November 7, 2014 58
Memory Task
• Goal: Test our hypothesis that lexical and contextual modifiers lead to different representations of alternatives, which may affect speakers’ ability to remember them later.
• We looked at how well participants recognized the alternative images versus how likely they were to misremember a similar image that they have never seen.
Danielle Moed - University of Toronto - November 7, 2014 59
Signal Detection Theory
Danielle Moed - University of Toronto - November 7, 2014 60
Predictions
• Lexical modification leads to better memory– Alternatives are part of lexical meaning– Informativeness
OR
• Contextual modification leads to better memory– Negative property in modifier (e.g., … without
asparagus) = more informative
Danielle Moed - University of Toronto - November 7, 2014 61
Mean Likelihood of a Hit/False Alarm by Condition
Danielle Moed - University of Toronto - November 7, 2014 62
Mean Likelihood of a Hit/False Alarm by Modification Type
Danielle Moed - University of Toronto - November 7, 2014 63
Mean Response Time (ms) for a Hit/False Alarm by Modification Type by Condition
Danielle Moed - University of Toronto - November 7, 2014 64
DISCUSSION
Danielle Moed - University of Toronto - November 7, 2014 65
Goal 1
• To examine how speakers conceptualize a referent in an alternative set; first, by looking at how this conceptualization is reflected in their choice of a referring expression.
Danielle Moed - University of Toronto - November 7, 2014 66
What we found
• Lexical modifiers in the lexical condition
• Contextual modifiers in the contextual condition
• No preference for lexical VS contextual modifiers in the both condition
Danielle Moed - University of Toronto - November 7, 2014 67
The alternation between lexical and contextual modifiers
• Visual attention during language planning?
Danielle Moed - University of Toronto - November 7, 2014 68
Goal 2
• To examine how the type of modification speakers produced affected their memory for an unmentioned alternative.
Danielle Moed - University of Toronto - November 7, 2014 69
What we found
• Crucially
Memory was better when individuals produced a contextual modifier versus a
lexical modifier (in the both condition)
Danielle Moed - University of Toronto - November 7, 2014 70
Why do contextual modifiers better memory?
• Informativity?
• Non-linguistic phenomenon?– Visual Attention
Or is it something else?
Danielle Moed - University of Toronto - November 7, 2014 71
Representation of language affects memory?
• Informativity of referring expressions
• How people conceptualize relationships between entities how informative those relationships are
Danielle Moed - University of Toronto - November 7, 2014 72
REFERENCES• Brady, T.F., Konkle, T., Alvarez, G.A., & Oliva, A. (2008). Visual long-term memory has a massive storage capacity for
object details. PNAS, 105 (38), 14325-14329. • Brennan, S.E. & Clark, H.H. (1996). Conceptual pacts and lexical choice in conversation. Journal of Experimental
Psychology: Learning, Memory and Cognition, 22, 1482-1493. • Cohen, A. (1999). How are Alternatives Computed? Journal of Semantics. 16 (1), 43-65. • Evans, V. & Chilton, P. (Eds.) (2009). Language, Cognition and Space: The State of the Art and New Directions. Advances
in Cognitive Linguistics. London: Equinox Publishing Company. • Fraundorf, S.H., Watson, D.G., & Benjamin, A.S. (2010). Recognition memory reveals just how CONTRASTIVE contrastive
accenting really is. Journal of Memory and Language, 63, 367-386. • Grice, H.P. (1975). Logic and Conversation. In Cole, P. and Morgan, J. L. (Eds.) Speech Acts. New York, NY:
Academic Press. • Heller, D. & Chambers, C.G. (2014). Would a blue kite by any other name be just as blue? Effects of descriptive choices on
subsequent referential behaviour. Journal of Memory and Language, 70, 53-67. • Herskovits, A. (1987). Language and Spatial Cognition: An Interdisciplinary Study of the Prepositions in English. Cambridge,
MA: Cambridge University press. • Gotzner, N., Spalek, K & Wartenburger, I. (2013). How focus particles like ‘only’ hamper the rejection of contrastive
alternatives. Poster presented at the 26th Annual Meeting of the CUNY Conference on Human Sentence Processing, Columbia, SC.
• Kennedy, C. (2007). Vagueness and grammar: the semantic of relative and absolute gradable adjectives. Linguistics and Philosophy, 30 (1), 1-45.
• Kennedy, C. (2012). Adjectives. In Russell, G. and D. Graff Fara (eds.), Routledge Companion to Philosophy of Language. Routledge.
• Kennedy, C. & McNally, L. (2005). Scale structure, degree modification, and the semantics of gradable predicates. Language, 81 (2), 345-381.
• Macmillan, N.A. & Creelman, C.D. (2004). Detection Theory: A User’s Guide. Mahwah, NJ: Lawrence Erlbaum Associates, Inc., Publishers.
• Sedivy, J., Tanenhaus, M., Chambers, C., & Carlson, G. (1999). Achieving incremental semantic interpretation through contextual representation.Cognition, 71, 109-147.