These are questions that students have asked, from past years as well
as from the current year.
Q: I was just wondering if the grades at the end of the course
will be curved.
A: Well, "curving" means that letter grades are determined by means and
standard deviations (based on something called a "normal
distribution"). I don't do that. As mentioned on the page
that reports the grades, I can only finalize the scale (which
determines who gets an A, B, etc.) at the end of the quarter, once all
grades are in. (You wouldn't want me to decide now that a 90% is
necessary for an A, only to end up with 5 students receiving more than
90%, would you?) For purposes of planning, etc., you can assume
that 90% and higher will be some kind of A, 80% and higher will be some
kind of B, and so forth, but I reserve the right to use a slightly
different scale if it's necessary.
Q:
> I was wondering if studying the pages on the lecture outlines
would
be
> enough and if you lectured outside of what was outlined.
A:
Nope, I only test out of the pages that are listed in the outlines
(which
are also listed on the respective "Midterm Info" page).
Q: How do I add the class?
A: See the staff in the Student Services office, room 1533 of the
McGill Hall Annex. It's open 9-12, 1-4 M-F.
Q: I have the Peter Gray Psychology book, but it is the
third (or fourth) edition. Will that be sufficient for this
course, or do I
need to get the fifth edition?
A: It's dangerous. I don't exactly know what's in previous
editiosn, but there could be changes that end up being
important. So... The answer is "do so at your own
risk." If I were in your position, I'd probably get a hold of the
most recent edition.
Q: Is it that we have to read the entire chapter as
assigned by the syllabus or just the pages you have written on website
with the corresponding lecture?
A: You're only responsible for the page ranges listed on the
website in the "Lecture Outlines" section, and in the information page
for the corresponding test (which should list the same page ranges).
Q: For the test, should we focus more on the class notes or
will the majority of the test be from the text book? Also, what
would
you say is the best way to study? Would you suggest using the
text
book more or the notes?
A: Here's the way I make up the test: It's mostly based on
material from lecture, although I try to make the material consistent
with the book too. There could be questions
on
those topics that I've talked about where there isn't anything in the
book.
So, the best strategy in my opinion is to begin with the class notes,
and know them decently. Then, study the stuff in the book that
corresponds to what I talked about in class. That way, you have
two perspectives on any given topic -- the stuff that I have to say,
and the way that Gray says it.
Add to that, of course, the parts of the textbook that I specifically
told you to know about -- anywhere where I said in lecture, "and know
about pages..." That stuff is completely fair game.
Q: I was wondering, which scantron sheet do we need to buy for
our upcoming midterm??
A: The pink one -- Scantron F-1712-ERI.
Q: I went online and looked at your site and saw the pages
and such under the heading "midterm." I was wondering if the questions
that you put under that heading were ones that I should be doing, or
just looking over
to see if I know.
A: The focus questions that I list on the midterm pages are for
your studying convenience. Nothing exactly like those questions
will appear on the midterm or the final, but those questions are a good
way to study. If you read and study the book, and study your
lecture notes, you should be
able to answer the focus questions, which will indicate that you
probably know the material reasonably well.
Q: It seems like a lot of what we've been covering (Higher
Order Processes) isn't in the book, or at least you didn't list many
pages
to cover on the Lecture Outline page. Will there be more reading from
the
book or is this stuff only covered in class?
A: There will be more reading from the book generally -- I'm
about to put up the readings for the development section. But
much of the stuff on knowledge representation is mostly not in the
book.
Q: I didn't quite understand the definitions of "predicate"
and "argument" today. Can you give me some
definitions?
I understood the examples mostly.
A: What we'll see next class is that an argument is a 'concept'
(and we'll spend time discussing conceptual structure). So, with
that in hand, here are some definitions from a different text book:
A concept is a class or category that subsumes a number of individual
instances [so, this is basically what an argument is]. An
important way of relating concepts is through propositions, which make
some assertion that relates a
subject (e.g., chickens [that's the argument, or the concept] and a
predicate (e.g., lay-eggs). (Gleitman, Fridlund, & Reisberg,
Psychology).
So, propositions make statements -- they attribute properties to
arguments [=concepts]. The process of attributing a property to a
concept is called
'predication', and the part of the statement that makes that
attribution is
the predicate.
And, as a last ditch attempt, here are definitions from Merriam-Webster
dictionary (http://www.m-w.com/) --they're pretty good. The noun
'predicate'
is defined as:
1 a : something that is affirmed or denied of the subject in a
proposition in logic b : a term designating a property or relation
2 : the part of a sentence or clause that expresses what is said of the
subject and that usually consists of a verb with or without objects,
complements, or adverbial modifiers
[in both of these cases, the word 'subject' means what we mean by
'argument' or 'concept']
Here's the definition of the verb:
2 a : to assert to be a quality, attribute, or property -- used with
following of <predicates intelligence of humans b : to make (a term)
the predicate in a proposition
Q. A question I have is that about predicates (1,2,3 place
predicates). I'm not exactly sure what it means but if I say that
:
1place predicate = birds can fly
2place "
"
= birds can fly and chirp
3 place "
" =
birds can fly, chirp, and lay eggs
would these be true statements as predicates?
A. No, not quite. Your example of a two-place predicate is more
like two one-place predicates, like can-fly(birds) and
can-chirp(birds). A two-place predicate is one that MUST have two
arguments, or else it wouldn't make any sense. So, think of any
relationship that is between two things -- that'll be an example of a
two-place predicate. So, "left-of" is a two-place predicate,
because for something to be "left-of," it's got to be left-of something
else. "mother-of" is another example.
> What is the difference between classical conditioning and operant
> conditioning?
>
Classical conditioning is about associating new stimuli to
already-existing reflexes. Dogs come to the world such that they
salivate when they get food -- that's an already existing reflex.
What Pavlov did with his dogs was associate a new stimulus -- a bell --
to that already existing reflex -- salivating when they get food.
Operant conditioning is about acquiring new ways of operating on the
world. Thorndike's cats learned a new way to operate on their
environments -- when they pulled a lever, they were released from the
puzzle box.
So, a big difference between classical and operant conditioning is that
with operant conditioning, the response is new (cats don't come to the
world such that they know how to pull levers to get out of puzzle
boxes), whereas with classical conditioning, the response is not new
(dogs already know how to salivate), but the animal comes to exhibit
that response in a new situation.
> What is an operant response?
>
Thorndike's cat pulling a lever to get out of a puzzle box is an
example of an operant response. So too is a rat in a Skinner box
pushing a button to get a food pellet.
> Also, what are the differences between generalization,
discrimination
> training, and disciminative stimuli?
>
Generalization: A dog is classically conditioned by pairing one
kind of bell to food, such that eventually, when that bell is rung, the
dog salivates. Now you ring a different bell, say, one that rings
with a slightly lower pitch. When you ring it, the dog will
nonetheless salivate. This illustrates that stimuli that are
similar to but not exactly the same as the stimuli that have been
paired with an unconditioned reflex will themselves elicit that
unconditioned reflex.
Discrimination training: This is a combination of generalization
and extinction. Take the above example of generalization, with
the lower-pitched bell. As mentioned, because of generalization,
an animal will salivate to the lower-pitched bell, even though it
wasn't the exact bell the animal was trained with. Now, you
continue training, pairing the original higher-pitched bell with
food. At the same time, you pair the new, lower-pitched bell to
NO food, thereby extinguishing the generalized response to that
lower-pitched bell. Eventually, the animal will salivate only to
the original higher-pitched bell, and not to the new lower-pitched
bell. This shows that the animal can in fact distinguish the two
kinds of bells.
Discriminative stimuli: Imagine an operant conditioning situation
-- Thorndike's cat. When the cat pulls a lever, it gets out of
the puzzle box. But, this only happens when a red light is on
inside the puzzle box. When the red light is off, the lever does
nothing. The cat will come to learn that it can only pull the
lever when the red light is on. Thus, the red light is a
"discriminative stimulus" -- it is a stimulus that indicates that a
particular operant response will be effective.
> Is chainig similar to shaping? If not what are the differences
or
> examples?
>
Shaping is like the "you're getting warmer" game (when your friend is
trying to guess something or find something, and you say "you're
getting warmer..." to indicate that they're getting closer and
closer). Shaping involves reinforcing successive approximations
of a target operant response until that target response itself is
emitted. Shaping is intended to allow an animal to learn a
low-likelihood response. So, imagine you want to train your dog
to roll over on its back. To use straight operant conditioning,
you'd have to wait a million years for your dog to roll over on his
back on his own, so that you can reinforce his response. What you
can do, though, is train your dog with "you're getting warmer" --
chaining. First reinforce the dog just for sitting. Once
the dog does that, now you reinforce him for lying down. Once he
does that, then you reinforce him for lying on his side. Finally,
you only reinforce him once he's lying on his back.
Chaining is different. Take the example above of a discriminative
stimulus, with the cat that can pull a lever only when a red light is
on. This red light is obviously important. In fact, it's
important enough that the cat will do things to turn on the red light,
if it can. Imagine that when the cat pushes a button with her
nose, it turns on the red light. Once the red light is on, the
cat can pull the lever and get out of the box. Eventually, the
cat will press the button to turn the light on, so that she can pull
the lever to get out of the box. That's chaining. Every
time you spend money to get something you like, it represents the end
of a chaining sequence. Why? Because you did some response
-- worked at your job -- in order to get money, so that you could do
some other behavior that involves spending the money to get something
reinforcing.
> I dont really understand Sperling's partial report procedure.
In the full-report condition, people saw all 12 letters, and could
report about 4 or 4.5 of them. Why did this happen? Is it
because they actually saw all 12 letters (i.e., all 12 letters got into
iconic memory), but only 4 (or so) were reported before the letters
disappeared from iconic memory? Or is it because people only
stored 4 (or so) letters in iconic memory in the first place, and
that's all they could report?
In partial report, people still saw all twelve letters, but were only
asked to report a RANDOMLY determined set of four of those letters (the
top row, the middle row, or the bottom row). People were
basically able to report all of those four letters (3.5 or
so...). Because people could recall nearly all of a RANDOMLY
determined subset of letters, that must mean that in principle, all 12
letters were available for reporting. If only 4 of the 12 letters
were available at all, then people should only be able to report a
random four of those letters one-third of the time.
So, summing up, when I show you a bunch (12) letters really quickly,
you can only report about 4 of those. That seems to be because
you see all 12 of the letters, but they fade so fast from memory that
you can only report four of those letters before they're gone.
> What was the point of the Craik & Tulving experiment?
Did it show that maintenance rehearsal doesn't lead to encoding?
>
The Craik and Tulving experiment wasn't so much about maintenance
rehearsal as it was about shallow processing (thinking about the
superficial properties of things -- how they look, or only very simple
aspects of meaning) versus deep processing (thinking extensively about
what something means, how it relates to other things, etc.). The
Craik and Watkins experiment ("tell me the last word on the list that
starts with 'G'") was much more about the ineffectiveness of
maintenance rehearsal.
> I am completely mystified as to what the Cooper/Shepard was.
>
So, I show you a letter 'R,' and ask you a question about it -- is it a
regular letter 'R,' or is it mirror-reversed (as if you were seeing it
in a mirror). Trick is, the regular- or mirror-reversed-letter is
not upright; it's rotated away from upright varying amounts. The
cool result here is that the time it takes you to answer the question
(regular? or mirror-reversed?) is directly proportional to how rotated
the letter is -- the further the letter is rotated from upright, the
longer it takes you to judge whether it's regular or
mirror-reversed. This strongly suggests that what you do to
answer the question (regular or mirror-reversed) is look at the letter,
and rotate it in your mind until it is upright. Once it is
upright, you can judge whether it is regular or mirror-reversed.
In turn, this shows that people take longer to rotate letters that need
to be more rotated. This is the essence of an analogical
representation -- it behaves in your mind just like it would behave in
the real world.
> I know what a concept is, but what is the difference between a
formal
> and natural concept? Does this have to do with the two differing
theories
> on concept structure that we studied, i.e. the one involving a
more fixed
> set of rules in the definitions, and the prototype theory that
allowed for
> "close resemblances" to fit under the categories of concepts?
>
Natural concepts are concepts that exist in the real world -- birds and
chairs and trees, etc. Formal concepts are concepts that we as
human beings have defined. Examples of formal concepts are things
like 'odd number' or 'uncle.' An odd number is something that we
have defined as any number not evenly divisible by two. An uncle
is something that we have defined as the brother of one of your parents.
We brought these up to make a couple of points. In principle,
formal concepts _could_ be described by the definitional theory --
they're defined by a definition, after all. But they're still
not! People think that some odd numbers are better examples of
odd numbers than other odd numbers, and people think that some uncles
are better examples of uncles than other uncles. So, even formal
concepts, which _could_ be described by the definitional theory, are
not. Prototype theory wins.
> I understand the Stroop effect, especially through the experiment
on
> color words, but I want to know, what are the positive or neagative
> effects of automaticity (adv/disadvnatages). I'm also a little
sketchy on
> the macleod and dunbar experiemnt.
>
Automaticity's advantages are self-evident, based on its properties
(fast, accurate). The disadvantage of automaticity is that,
because it's obligatory and inflexible, automatic processes kick in
even if you don't want them to. So, when you wanted to pick
something up on the way home from school, but ended up at home instead
without running the errand, it's because an automatic (obligatory,
inflexible) process took over and prevented you from doing something
flexible -- going somewhere on the way home.
> Is the diagnosis of automaticity determined by the illustration
> (through the experiment of color words, for example) of one
process that
> we cannot stop voluntarily, and another process that we intended
to do
> (intended vs uninteded?)? What practical purpose does it serve?
The book
> mentions how the stroop effect can be used to determine literacy in
> children by their vulerability to the stroop effect.
>
Right. The idea is that if an unintended process interferes with
the performance of an intended process, then the unintended process
must be relatively more automatic. (If the unintended process had
been less automatic than the intended process, then the unintended
process would have not have gotten involved, and so interference would
have resulted.) By this logic, you can use the color-word Stroop
effect to diagnose how automatic reading is.
> Would a map be possibly considered as a analogical representation?
> Techically, it is not arbitrary, and although it doesnt match the
visual
> images exactly, it does mean something and have a relationship
with the
> nature of the thing it stands for.
>
Yup, a map (as a mental representation) would definitely be considered
an analogical representation.