LING 810 Chapter 6: Week 6 - Lupyan, 2008
LING 810 Week 6 - Lupyan, 2008
From Chair to “Chair”: A Representational Shift Account of Object Labeling Effects on Memory
Abstract:
● How does calling things by a name affect memory?
● Memory was worse when objects were classified using a category name
● Authors claim a “representational shift account”, meaning that the typicality of the category
compared with the item being categorized affects whether the item is remembered or not
○ Atypical items are less likely to be remembered
● Results suggest that “labeling a familiar image distorts its encoded representation”
Introduction
● Words often denote entire categories
○ I.e. table does not mean a certain table in particular
● Goal of this study: “investigate the consequences of labeling familiar items with their
category”
○ Different from studying the meaning of words and how those relate to
represent/remember objects
● Carmichael et al. (1932) had participants redraw a shape similar to X after the original image
was called either an hourglass or a table
● This study asks: What is the consequence of using a name to classify familiar objects?
○ No consequence - names are just a label for things we know about
○ There is a consequence - object names highlight the relationship between items and
their categories; an object that has its category named will be influenced by other
members of that category
● Knowing what an object is and being able to name it are two separate processes just as
categorizing and naming an object are two separate processes
● Using an object’s category name may activate top-down category information association
○ Thus, category names may be able to provide feedback to alter the representation of
an item
● “Words draw our attention to object categories”
○ Previous findings: Children pay attention to objects that are named, naming items can
help differentiate them, calling things the same name prompts children to look for
similarities between the items, etc.
● Labeling items has a learning effect: Labeled categories are learned faster than unlearned
categories
● Single-item recognition model:
○ Global matching retrieval process - new items are matched with existing items in a
category; the closer the match, the better the item will be recognized
○ More exposure to an item means more of the item’s features are stored; improves
recognition
● Hypothesis: familiar items that are labeled will be recognized less than familiar items that are
not labeled
○ Representational shift account
● Studies with vision have shown top-down activation effects which affect perception
● False recognition studies: “false recognizing novel items as old”; occurs because the features
encoded are shared by many category members
○ Is predicted to result in a high number of false alarms (i.e. saying they’ve seen an
item/picture before when they actually haven’t)
○ Other studies have shown that categorical coding study contexts results in worse
memory of items compared with study contexts that enhance item-specific features
■ E.g. studies with pictures of cats, pictures integrated in scenes vs. physical
features of those pictures
● Representational shift is different from false recognition in that it predicts low hit rates for
labeled items because the stored representations don’t match test stimuli
○ I.e. participants will often not recognize test stimuli because the label name distorts
the representation of these stimuli by activating top-down effects
○ “Also predicts that the effect of classification on memory should be mediated by the
typicality of the item being overtly classified (i.e., labeled)”
■ Ambiguous items should be more affected by labeling
● The experiment used familiar objects that already have high semantic associations
○ Participants asked to either classify items or make a preference judgment (which does
not have to do with object category)
■ Participants are likely to implicitly categorize objects anyway, but effects
should be stronger when participants are overtly classifying the object
● 6 experiments total
Experiment 1
Method
● Participants: 18 participants
● Materials: 40 pictures of chairs and 40 pictures of lamps taken from an IKEA catalog (20 of each
for the study session and 20 of each for the test session); each picture was matched with a
critical lure to produce 80 pictures in total
○ Presented on a screen to participants
○ Study sessions: category responses and preference judgments
○ and test sessions: classify whether a picture was old or new
○ Each picture would also have a matched critical lure, which would be similar to the
original, but different in a distinct way
Procedure
● Collected typicality ratings for the experiment stimuli from 10 individuals
● Study phase:
○ Classification block: Categorize pictures either as chair or lamp
○ Preference block: indicate preference for the objects
○ Pictures presented quickly and each one presented twice
○ 8 blocks, 10 pictures each, alternating classification and preference blocks
○ Every item seen in both contexts across all participants
● Test phase:
○ Participants told they would see more pictures, some of the same as before but some
similar, yet different
○ Stimuli fall into four categories:
■ Old seen in classification
■ Old seen in preference
■ Lure similar to a picture in classification
■ Lure similar to a picture in preference
○ Different types of pictures were given all together (not in blocks)
Results
● Study phase:
○ Overall accuracy was high; errors were likely due to short response time, vagueness of
objects, and a small proportion of errors were due to invalid responses (classifying
when they shouldn’t, or indicating preference when they shouldn’t)
○ More typical items classified more quickly and accurately
■ More typical items were also better liked
○ Likability did not correlate with reaction time or recognition memory
● Test phase:
○ Recognition memory above chance
○ “Participants had lower recognition memory for items they had classified compared
with those for which they had indicated preference”
○ “No significant difference between the lures most similar to the classified items and
the lures most similar to the items for which preference had been indicated”
● Typicality effects at test:
○ Typicality ratings correlated with recognition memory
○ Participants had a higher hit rate for atypical items compared with typical items
■ More typical items had slightly higher false alarms