Q2: Lecture 11 Flashcards
Endel Tulving
one of the best known and highly esteemed of all cognitive psychologists. In 1972,
Tulving addressed an auditorium full of his colleagues at a conference held in Toronto. During his
address, Tulving posed a rather simple question. He asked the cognitive psychologists gathered
there to indicate what they thought were the most interesting questions about memory.
semantic memory
cognitive
psychologists were interested in abstract factual memory – something we call semantic memory.
episodic memory
memory for events or life episodes – experiences that occur at some specific time in some
specific place. If I ask you what you were doing yesterday at noon, or I ask you to tell me about the
events you experienced on your last birthday, or I ask you to recall the last sentence I spoke, I am
asking you to tap into episodic memory. Because most memory research up to the time of Tulving’s
address focused on people’s ability to recall lists of words, the emphasis was clearly on episodic
memory
conceptual memory
When we talk about semantic memory, we are talking about conceptual memory. Concepts are units
of meaning, abstract ideas, and general notions about things. Our semantic memories are made up of
vast numbers of concepts - concepts like BOOK, LOVE, HOME, PSYCHOLOGY, WORK, FUN, and
on and on.
concept
Concepts are units
of meaning, abstract ideas, and general notions about things. Our semantic memories are made up of
vast numbers of concepts - concepts like BOOK, LOVE, HOME, PSYCHOLOGY, WORK, FUN, and
on and on.
category
Any concept that is broad enough to include several
other concepts is considered a category. The concept DOG is a category because it includes many
specific types of dogs. The concept GERMAN SHEPHERD is a category because it includes many
specific German Shepherds.
superordinate concept
broad and correspond to large categories; superordinate concepts are also categories, this is not the case for subordinate
concepts.
subordinate concept
concepts are more specific and can correspond to smaller categories; Highly specific concepts (e.g., YODA, HALLOWEEN, and BILL BURR) are described as
subordinate but are not categories because they do not subsume any other concepts
traditional concept-learning
Each trial of a traditional concept learning study involves the presentation of a visual
stimulus. (Some examples are shown to the left.) The learner’s task is to determine
what concept or category the experimenter has in mind. In response to each stimulus,
the participant can say either “Yes” (I think that stimulus is an example of the concept)
or “No” (I don’t think that stimulus belongs in the to-be-learned category). After each
response, feedback is provided about the learner’s accuracy, and then the next trial
begins with the presentation of a new stimulus. This continues until the participant
meets an arbitrary learning criterion, like 10 correct responses in a row. Once the
criterion is met and the concept is considered learned, a new set of trials begins, and
the participant attempts to learn a different concept.
artificial category/concept
The rules that define
these artificial concepts are described in terms of the stimulus dimensions used to create the test stimuli.
learning criterion
like 10 correct responses in a row. Once the
criterion is met and the concept is considered learned, a new set of trials begins, and
the participant attempts to learn a different concept.
stimulus dimension
A stimulus dimension that is critical to the concept definition is a relevant dimension,
whereas those that are not part of the concept definition are irrelevant
dimension value
relevant dimensions
A stimulus dimension that is critical to the concept definition is a relevant dimension. For example, If
the to-be-learned concept is RED THINGS, then Color is a relevant dimension
irrelevant dimension
whereas those that are not part of the concept definition are irrelevant. Shape and Size are irrelevant dimensions
unidimensional rule
there is only one dimension that is relevant, and the learner
merely needs to discover the concept-defining value for that dimension to be successful.
conjunctive rule
stimuli are only members of the to-be-learned category when they
are both small and blue, so the to-be-learned CONCEPT 2 is SMALL BLUE THINGS. For this
concept, Color and Size are relevant dimensions and Shape is irrelevant. This concept is defined by a
conjunctive rule because it involves the intersection (or conjoining) of values on two different stimulus
dimensions
disjunctive rule
category members are always red and/or triangles, so the to-
be-learned concept is RED OR TRIANGLUAR THINGS. Here, Color and Shape are relevant, and
Size is irrelevant. In addition, this concept is defined by a disjunctive rule because stimuli are
category members if they exhibit either of two different features, but the presence of both features is not necessary
inclusive-or rule
examples
that exhibit the intersection of the critical features are included in the category
exclusive-or rule
stimuli that exhibit the
intersection of the relevant dimension values are excluded from the category
Eleanor Rosch
a cognitive psychologist interested in concept formation and conceptual knowledge.
She recognized that the traditional concept learning research suffered from its exclusive
concentration on artificial concepts. Rosch realized that many of the concepts and categories about
which we learn cannot be defined in terms of highly specific rules.
natural category
natural categories
like DOG, TREE and INSECT are difficult or even impossible to define in precise terms, but we
nevertheless learn about and understand these concepts.
Rosch pointed out that natural categories are often made up of ill-defined concepts, and rather than
having definite boundaries, membership in a natural category can be a matter of degree rather than
of absolute belongingness or non-belongingness. Moreover, she thought color categories were an
especially good illustration of the nature of natural categories.
ill-defined concept
natural categories are often made up of ill-defined concepts,
Labov (1973)
provided an excellent illustration of the
nature of natural categories. He conducted a simple experiment in which
people saw a series of drawings of objects and were asked to name
each one.
fuzzy boundary
gray area as the fuzzy
boundary between the two natural categories.
context sensitive
he fuzzy boundary between the
two natural categories is context sensitive and can shift toward or away from a concept depending on
the specific situation.
object recognition
a simple form of perception. That is, object recognition involves processing
sensory inputs so that an understanding of the object-related sensory qualities is established
image-based theory
assume that each object is represented in memory as a
sensory pattern and an associated semantic label. The internalized, mental representation of the
sensory pattern is called an image. When we encounter the term “image” we usually think of visual
images, but psychologists use the term in a more general way. Thus, although we can use the term to
refer to a visual image, we can also use it to refer to a sound image or a tactile image; we do not limit
the use of the term to the visual sense modality
template theory
The prime example of an image-based theory of object recognition is template theory. A template is a
mental image that corresponds to a copy of the sensory pattern resulting from experience with an
object
Palmer, Rosch, & Chase (1981)
canonical perspective
The researchers noted two
important things about peoples’
ability to identify the objects in the photographs. First, for any one object there was usually one perspective that was associated with the easiest object recognition. Palmer et al. (1981) called this the canonical perspective
parts-based theory
A second type of object recognition theory suggests that we recognize entire sensory configurations
by first identifying simpler elemental parts of the configuration. These simple but meaningful elements
are called features. Features are basic, indivisible perceptual primitives, like the simple shape of a
circle, a right angle, or a solid color. According to parts-based theories of object recognition, the
perception of a whole first requires the detection of the featural parts that make up the whole. For
example, we might be able to recognize a horse by first identifying features like “long face”, “muzzle”,
“mane”, “four hooves”, “four fetlocks”, and “full distinctive tail.”
feature
are basic, indivisible perceptual primitives, like the simple shape of a
circle, a right angle, or a solid color
Hubel & Wiesel
conducted extensive
studies designed to identify the receptive fields of nerve cells that process visual signals; found that neurons at specific levels of visual information processing
exhibited similar receptive fields, and that as processing progressed from one level to the next, the
receptive fields got more and more complex
receptive field
the receptive field is the sensory pattern to which a neuron is
specifically tuned, such that it becomes highly active whenever that stimulus is encountered by the
organism
spot detector
the neurons of the retina and lateral
geniculate nucleus of the thalamus that support the earliest stages of visual processing are
associated with very simple receptive fields. Neurons in these areas respond maximally to spots of
light of specific sizes falling on specific parts of the retina. These neurons, therefore, are often called
spot detectors.
edge detector
some
neurons in primary visual cortex act as edge
detectors, because they fire best in response to
bars of light in specific orientations. For example,
some of these edge detector cells fire best when a
vertical line is presented, others when a horizontal
line is presented, and others when diagonal lines
are presented.
RBC
One of the best-known feature-based theories of object recognition was
proposed by Irving Biederman (1987, 1990). He calls his theory the recognition-
by-components (or RBC) theory because it suggests the key to object
recognition is successful identification of the constituent parts or components of
an object.
Geon
Object components are perceptual primitives, and Biederman attempted to identify the minimal set of perceptual primitives that are required to support object recognition. He called these object features geons, and each geon is a basic three-dimensional shape, like a cylinder, a rectangular solid, or a cone.
recoverable drawing
All the degraded drawings were erased to the same degree, but the specific
portions that were erased varied. For some, he erased parts that he guessed were not essential to the identification of geons. He called these the recoverable drawings (R; center column) because he intuited that
people could recover the component geons despite the erasures
non-recoverable drawing
For other drawings, he erased portions that he believed would make it very difficult or impossible for people to recover the geons. He called these the non-recoverable drawings (NR; right-hand column)
Biderman & Blickle (1985)
provided even more
compelling support for the RBC theory; a more
objective definition of recoverability was employed. Biederman hypothesized that the vertices of a drawing are especially critical to geon identification whereas line
midsegments are less important.
locus of degradation
they also manipulated the location of the deleted lines
Davenport & Potter (2004)
They showed them photographs of scenes and
asked them to identify the object that was in the
foreground of each one. Thus, instead of using line
drawings, they used more realistic representations of
the to-be-recognized objects.
context effect
cannot be explained by purely data-driven theories
