Cognitive Psych - Exam 2

object based neglect

*refer to powerpoint notes •Objects that begin in the left field of space •Start rotating, neglect moves with object rotation •Don't see object in 1 because in bad side of space, don't see it as rotate •But would see object that started on right in 1, would see it on the left on box 4 •can't focus on object as rotates → regardless of location (whether on right or left side of visual field)

Visual Neglect: An Example

*refer to powerpoint notes •Visual neglect patients have extreme difficulty detecting targets on left •Stationary display •Visual neglect person doesn't notice the highlighted circle on left box 4 or extreme delay in noticing, when on right, no worries •Baseline experiment •Highlight left on stationary display and could not detect it

Attention Can Spread Throughout an Object (But Not Across Two Objects)

*refer to powerpoint notes •What people attend to not to just a spatial location but an object •Present cue → people see 2 bars, fixation point between 2 bars, cue one of 4 corners, cue in this case is the top right •Present stimulus is one of 4 corners, measure reaction time to detect stimulus -80% of the time cue is same as stimulus -how does inaccurate cue affect reaction time to find target stimulus? • Pretty fast when cue and target are same • A bit slower when move across object but different location (point b) • If move same amount of distance but cross to different object, point c, even slower • B is faster than c. B gets more attention than c • Attend mainly to a. but → Attention is spreading within object, from A to B •Is attention being to spatial location or the object (right rectangle)? -a little bit of both -same distance but other rectangle © -same object but different location (B) -only small difference between location b and location c, directed to attend to a, (faster at a than b), but faster at b than c. so spotlight is spreading within object (from a to b) but not across 2 objects (from object A to C)

Object-based and Location-Based Neglect

*refer to powerpoint notes •don't see the left hand side of space nor attend to it. •any object that begins on left side of space, going to lose as rotate to other side of space, still can't see where object highlighted on right •what was on the left and is now on the right still cant see → neglect travels with object rotation (object neglect) •can't see left side of field (location neglect)

Location based neglect

*refer to powerpoint notes •shaded area is what they have trouble attending to → anything on left, cant see, •if moves to the right, they can see it •rotate clockwise, can see in 3. And highlight in 4, they'll be able to see it. → in right •don't find this neglect alone in patients

Visual Neglect with Motion

*refer to powerpoint notes •½ patients no longer detected target on RIGHT -suggests object based neglect •½ patients no longer detected target on either side -suggests object-based and location-based neglect •rotate barbell, highlight one end, → started on left, now on right. Could not detect highlighted right end •If on left, could detect it. If on right could not. On stationary display, left was problem. Now on rotation, right is the problem •detecting objects that first appear on right and moves to left, its okay.. because starts on right, → inability to see objects that first appear on their left field of space, And -general neglect of left field of space -so if object begins on right visual field, when moves to left, no longer see it → suggest have both object and location based neglect •no body shows just location based neglect

Examples of late selection: •Semantic context effects -Treisman (1964)

-Bilingual observers used in experiment -Unattended message was same as attended message (shadowed ear), but in other language and offset to the attended message •Left ear hearing same thing as right ear, just in different language in one ear -Gradually decreased lag between the two messages -At sufficiently short delays, observers noticed that the two messages were identical •SEMANTIC INFO LEAKS THROUGH THE FILTER •Another example of where information about meaning becomes available on a to-be-ignored channel comes from another experiment by Treisman, using French-English bilingual listeners. •She again used a shadowing task, instructing listeners to repeat back what was said on one ear or the other. •The message played to the shadowed ear was a passage in French from Charles Dickens' A Tale of Two Cities. The message played to the non-shadowed ear was the same passage but 1) read in English rather than French, and 2) delayed by several seconds relative to the to-be-shadowed message. •Treisman gradually reduced the delay between the two passages. --> •As she did so, listeners noticed that the message played to the two ears was the same message, just in different languages. --> •Information about meaning is clearly getting through the filter, otherwise how could the listeners recognize the two passages as being the same. Said differently, we can extract the meaning of a stimulus before we decide to ignore it. --> pattern recognition BEFORE selection

Examples of late selection: -Cocktail party phenomenon and your name (Moray, 1959)

-Recognize name on unattended channel --> somehow hearing your name is getting through channel and is not being filtered out --> going through pattern recognition Soon, evidence began trickling in suggesting that there were some problems with the idea that all selection was early selection. -Suppose that you are at a party, with many simultaneous conversations going on. You are in one corner, talking with several other people. Using spatial location as the channel, you hear very well everything that they are saying while drowning out—not letting through your filter—any sounds coming from the rest of the room. You are ignoring all but your own conversation. Now someone from the other side of the room, in a conversation that you have been successfully ignoring, speaks your name. You will hear it (and you will turn towards the person who spoke it). -Somehow, it got through your filter and you recognized those spoken words as your name—i.e., that particular stimulus made it through pattern recognition. -But how did your brain know to let that stimulus through pattern recognition? It cannot know that that stimulus is your name before it goes through pattern recognition, since pattern recognition is what tells us that the stimulus is our name. The only logical conclusion seems to be that everything goes through pattern recognition, and it is only AFTER something goes through pattern recognition that we decide whether to attend to it or not. Sensory channels --> pattern recognition --> selection --> short-term memory

Testing Capacity Models: -Johnson and Heinz (1972) (cont.) •Results

-Results: RT to light •Control: •Group 0 (1 list): 370 sec •Group 1 (2 lists, gender): 433 msec •Group 2 (2 lists, semantics): 482 msec •interpretation -processing takes effort. -hence, Group 0 slower than control and Groups 1 and 2 slower than Group 0. -early selection on main task leaves more capacity for processing subsidiary task -hence, group 2 (late selection) slower than Group 1 (early selection) •The control group is the fastest. Group 0 took a bit longer to respond to the light because the shadowing task presumably takes some minimal amount of effort to perform, effort that is then no longer available to perform the light task. -Group 1 responded to the light more slowly than Group 0, because Group 1 did have to perform some selection, albeit at a relatively early level of processing, leaving even less capacity left over for the light task. -had to distinguish physical characteristics -make decision based on pitch of voice -early selection -Group 2 responded even more slowly, presumably because they had to process the word lists even higher up the chain than Group 1, consuming even more mental capacity, leaving even less left over to perform the light task. Had to let ever word go through pattern recognition first and recognize word's meaning before can answer -Note that Group 1 is essentially doing early selection, deciding what to attend to based on physical characteristics, whereas Group 2 is doing late selection, deciding what to attend to based on the meaning of the stimulus.

Attentional capacity

-There is not just a single pool of mental effort/attention which we draw on for all tasks -(there is visual, hearing, etc.) -there are, at least to a degree, multiple pools, each roughly corresponding to a different sensory modality -just as we have different pools of muscles we can draw on for lifting (e.g. left arm vs. right arm, -dont have just one muscle use for lifting, have multiple limbs. I can perhaps lift 100 pounds with my left arm and 100 with my right, and if i am currently lifting 50 with my left arm, I can still lift 100 with my right - what I'm lifting with my left arm does not subtract from what i can lift with my right) -plus an executive controller -a decision maker, for either early or late selection -which also has limited capacity -decides how going to divide up mental effort across different tasks. -e.g. in johnson and Heinz study, how much am going to devote to the light task and how much to the shadowing task. -or in Bavelier study, how much effort should i devote to those small circles and how much to that distractor on the side.

A capacity Approach to Attention:

-according to capacity models of attention: •attention= mental effort - how much mental effort can devote to a task •performing mental/cognitive operations requires effort •the supply of mental effort is limited: limited capacity -only so much we can do at once. -think of it as like mental strength •Interference/attention deficits occur when task demands exceed total available capacity, causing performance to degrade --> So long as everything that we are doing mentally is within the limits of the available effort or our available capacity, we are fine; nothing bad happens. But if we are doing mental operations that require more effort than what we have available to us, we start making mistakes, we slow down, we mess up. •The higher up the chain we have to process material, the greater is the presumed load --> as we go from the sensory store through filtering through pattern recognition, etc., the amount of effort consumed generally increases. --> If I cut a stimulus off at the Filter stage, I've expended relatively little effort on it. But if I let it through the pattern recognition stage, say, I've spent more effort on it and hence have less effort left over for other things. --> That's important—the more effort I spend on a given stimulus, the less I have left over to use with other stimuli. So, if I really want to focus on something, I'm probably going to have to do it at the cost of letting something else go. --> •To a large extent, people can strategically control how effort is allocated - --> can use mental effort in various ways - can control how use attention (at subconscious level) --> -If its something going to end up ignoring, → the earlier ignore it, the better (if possible before pattern recognition) •Not going to waste too much mental energy on it •If ignore it later, after pattern recognition → more mental energy wasted •The higher up the chain a stimulus goes, the more energy is wasted

Pre-attentive processing: -Subitizing

-example of pre-attentive processing -As increase objects from 1 to 4, time is constant -in counting number of objects -After 4, increase time •When people are asked to indicate how many objects are in an array of objects, they respond quickly so long as the objects are 3, maybe 4, or fewer. -Furthermore, so long as the number of objects is within this range (1 - 4), people's latency to perform the count is not affected by the number of objects present. This phenomenon is known as subitizing. •When the number of objects exceeds 4, however, then people become slower to do the count. More important, their reaction time increases linearly with the number of objects. •It is as if, so long as the number of objects is small, we can immediately apprehend their quantity without devoting any attentional resources to that task. As the number of objects grows, however, that pre-attentive process can no longer operate, and we need to count the objects individually, devoting attentional resources to that task.

Capacity Approach to Attention: -A problem

-notion of capacity as cause of attentional deficits is circular reasoning -The fact that people have difficulty performing two tasks simultaneously (as in the Johnson and Heinz study) is taken as evidence that mental capacity is a limited resource. The assumption that mental capacity is a limited resource is simultaneously used as an explanation for why people have difficulty performing two tasks simultaneously. Why do people have trouble performing two tasks simultaneously? Because they have limited mental capacity. How do we know that they have limited mental capacity? Because they have trouble performing two tasks simultaneously. Similar to the case for levels of processing, we need a measurement of the amount of mental resources used by a task that is independent of dual task performance. -need an independent measure of resources required by a task, separate from attention deficits -To an extent, but only to an extent, we can make some progress by arguing that processing meaning generally takes more resources than making physical discriminations, since processing meaning goes farther up the overall processing chain. -But we can only take this argument so far. As one of my two tasks, I could choose one that requires a very fine physical discrimination, such as discriminating the pitch of two tones very close together in frequency. We are likely to find that such a discrimination, even though it does not require the processing of meaning, is still very likely to interfere with other simultaneously performed tasks. The discrimination is difficult, it takes a lot of effort. So even though it is done at a relatively low level in the processing chain, it is still taking a lot of our mental capacity. -Don't have a good ability to determine the capacity of a given attentional task

Cost of inhibiting an automatic process - -Besner, Stolz, and Boutilier

-show stopping an automatic process can cause penalty -Stroop task: -colored word presented -task to name the ink color -either whole word was colored, or just a single letter -incongruent trials slower than congruent trials -but, effect was smaller when just the letter was in color -But, overall response times were longer in the letter condition than in the word condition -Apparently, participants cold inhibit automatic process of reading by focusing attention on a single letter -the longer latencies, though, suggest that there was a cost to doing that inhibition -congruent= word blue printed in blue ink -incongruent= word blue printed in red ink -An experiment by Besner and his colleagues illustrates that although it may be possible to inhibit an automatic process, doing so extracts a cost. -They compared performance in a standard Stroop task to performance in a modified Stroop task, using the same words. However, in the modified task, only a single letter in the word was in color (the remaining were all black on white). Participants had to name the color of that letter. Doing so reduced the size of the Stroop effect but raised overall reaction times (see the chart on the next slide). -One interpretation of the results is that the reduced size of the Stroop effect indicates that participants were successful in inhibiting the automatic response of reading the word, since they needed to direct their attention to a single letter. The overall increased reaction times in the letter task, though, suggests that inhibiting that automatic response cost the participants something.

Two Explanations for Semantic Context Effects- •Treisman's Attenuation Model •Late Selection Models -Deutsch and Deutsch (1965) Norman (1970) Shiffrin (1975) Schneider and Shiffrin (1977)

As a result of studies like these, two alternatives to early selection were introduced, Treisman's Attenuation Model of attention, and late selection models. We'll describe each of those in turn.

Characteristics of Automatic processing: -Posner and Snyder (1975)

Automatic tasks require no (or little) attentional resources -should not cause interference (i.e. cause a performance decrement) on a second task -should not be interfered with by a second task it is not under voluntary or strategic control -it is always (usually) initiated upon presentation of the appropriate stimulus -processing is obligatory -or there is at least a cost of inhibiting them -once initiated, it cannot be stopped or inhibited (we have difficulty shutting it off or stopping it in the middle) -there is no intention to perform the process or conscious control -Ex. it is very hard if not impossible to look at a printed word and not read it It is not open to conscious awareness -we cannot introspect about who we do the process -we are aware that we are doing them but how we do them is not something that is generally open to conscious awareness

Selection and Task load: Experiment by Bavelier et al.

Discrimination task between diamonds and squares - in which participants saw an array of 6 circles, one of which had either a diamond or square in it. - they pressed on button if it were a diamond and another if it were a square. -Distractor: out on the edge, another stimulus appeared, either a large square or large diamond. This stimulus was entirely irrelevant to the task; participants could safely ignore it and no harm would be done. either compatible or competing distractor -4 different types of displays - -low load condition- a figure (either a diamond or square) appears in only one of six circles. -its low load because only one circle needs to be examined. -High load condition- a geometric figure appears in each of 6 circles, but only one circle has either a diamond or a square (target response) . -its high load because the person must search through all 6 circles to find the target response. 2 compatible distractor conditions: -low load and high load with compatible distractor condition -in which, the large distractor off to side is the same as the target shape - -reinforcement to correct response 2 competing distractor conditions: -low load and high load with competing distractor condition -in which, the distractor off to the side is different from the target inside the circle. -Generally, people are slower to respond when there is a competing distractor than when there is a compatible distractor - interference with correct response

Simon and Chabris Study: -Gorillas in our Midst

Easy attentional task: -count number of passes by people dressed in black Hard attentional task: -count number of passes by people dressed in white -count number of dribbles by people dressed in white either: -high similarity: attend to those dressed in black -low similarity: attend to those dressed in white --> differ from color of gorilla, thus less likely to see it -Easy attentional task: -high similarity condition- likely to see gorilla -ignoring white, selecting team in black, attending to -Gorilla also in black --> pay attention to it --> process at higher level, - later selection -Hard Attentional task: -low similarity condition- -get rid of stimuli as fast as can --> early selection -less likely to process gorilla

Allocating our attention (Endogenous Control): •Early selection vs. •Late selection

The great debate in the attention literature concerns where in the processing chain - early or late - do we exert our attention (when under endogenous control) •logically, these two alternatives are not mutually exclusive; attention could be applied both early and late in the processing chain. •Historically, however, these two alternatives have been treated as mutually exclusive. •Early and late selection are defined in terms of the general information processing model - debate over when selection occurs - filter, --> selection (attention is getting rid of stuff we don't want to attend to and preserving what we want to) -early selection occurs early in the processing chain; late selection occurs later in the processing chain

Divided Attention

When we've been talking about attention, we've been focusing on the idea that when there is too much out in the world that needs to be attended to at the same time, our performance suffers—we either begin to make mistakes, or we slow down, or we do both. In terms of load or capacity theories of attention, we would say that our capacity has been exceeded. Still, are there cases where we can perform two relatively demanding tasks at the same time without performance suffering? -(can we divide our attention across multiple tasks?) Can we exceed our attentional capacity without penalty? -e.g. driving a car and carrying on a convo -yes, under ordinary circumstances, able to. -listening to a lecture and doing the NY times crossword puzzle? --> would probably result in unsuccessful divided attention -i.e. decrement in performance -like in the shadowing tasks we discussed --> because non-automatic processes, very hard to do the two simultaneously -Spelke: taking dictation and reading short stories -claims t have taught people to both take dictation and read stories at same time. -After lots of practice, she claims that doing the tasks simultaneously resulted in no more dictation errors than doing dictation alone. -Likewise, reading times and scores on a comprehension test were no worse when people took dictation at the same time that they read a short story than when they just read the story. Perhaps. But I've never come across a replication of Spelke's experiment, so be cautious. -To summarize, in some cases, though not all, we can in fact perform what appear to be two rather demanding tasks at the same time, without incurring a performance deficit. What determines when we are able to do this? The answer seems to be that one or the other of the two tasks must be able to be performed automatically. -at least one of those tasks has to be "automatized"

Automatic vs. controlled processing

controlled processing: -open to influence by strategies (e.g. we can decide how to go about solving any particular Sudoku puzzle; it's not a behavior that is triggered simply by the sight of a puzzle) -we can choose to do them or not to do them -require attentional resources -Automatic <---> controlled: -a continuum, not simply two discrete states -a process can be automatized to a certain degree (e.g. understanding speech) -a complex process can consist of some automatic, some controlled sub-processes (e.g. driving a car, returning a tennis serve) -some are highly automatic but something can happen and cause need for controlled processing --> when reading, rely on automatic processing, but odd text --> may need to use controlled processing •Automatic processing is contrasted with controlled processing. •Controlled processing is often (but not always) open to conscious awareness, in the sense that we are aware not only of what we are doing but how we are doing it (Think about solving a Sudoku puzzle.). -It is also open to strategic control—we can decide how to go about solving any particular Sudoku puzzle; it's not a behavior that is triggered simply by the sight of a puzzle. - Controlled processing also requires attention. If we tried to do two things at once which both required controlled processing, we would encounter a performance decrement of some sort (unless each used a trivial amount of mental capacity). •It is probably best to think of automatic and controlled processing as being the end points of a continuum, rather than as an either/or distinction. A process can be automatized to an extent, but not necessarily entirely. Some sub-tasks of the main task can be automatic (e.g., braking for a red light). It can interfere more or less with other tasks. It can be under voluntary control, but perhaps only with great difficulty. And it could be more open or less open to conscious awareness. -Consider comprehending the sentence, "The horse raced past the barn." We comprehend that sentence seemingly effortlessly (though we are not really conscious of how we do that), and when we hear that sentence it's hard not to comprehend it—we cannot listen to speech without comprehending it. It is very far towards the "automatic" end of our continuum. -Now consider the sentence, "The horse raced past the barn fell." Everything is going along swimmingly, just fine, with no mental effort, until we get to that word "fell." Now, everything breaks apart. Our automatic processing led us in the wrong direction, and now we have to kick in more controlled processing, to figure out what the sentence is telling us. And unlike the automatic piece (It is hard to hear "The horse raced past the barn" and not comprehend it), we can choose to do or not to do the controlled processing necessary to understand "The horse raced past the barn fell." We may work at it for a couple of seconds and then choose to give up and move onto something else. And while we do work at it, we may be conscious of testing various hypothesis concerning what the sentence means. And while trying to make sense of it, other mental tasks are likely to suffer—it requires attention. •Complex processes are frequently a mix of automatic sub-processes and controlled sub-processes. -Consider once again driving a car. We have automatic processes of braking when a light turns red or when we see a stop sign. We have an automatic process to stop before making a left turn if there is on-coming traffic. - But when something untoward happens—we skid on ice, a child runs in front of the car, on-coming traffic crosses the centerline—we are likely to fall back on controlled processing, where we have to think about what to do and actively and often consciously make a choice.

Thresholds in Treisman's Attenuation Model

different threshold within Treisman's attenuation model. -in particular graph, threshold is determined by the permanent importance of the word. --> some words such as your name, have such a low threshold that they will almost always be recognized even on an unattended channel. --> other words like "rutabaga" which are not very important to many of us, have a very high threshold. Thus, they will rarely be recognized on the unattended channel (and maybe sometimes even on the attended channel) --> words like "boat" have an intermediate threshold, and sometimes will and sometimes won't be recognized on the unattended channel. -context can temporarily lower a word's threshold, increasing the likelihood that it will be recognized on the unattended channel

Selection and Task load (cont.) -Experiment results

how much does reaction time go up when have competing distractor vs. compatible distractor condition? cost of competing distractor: -Low load condition: reaction time increase by +30 msec when competing distractor compared to compatible distractor -Since low load task, not using all mental effort, left over mental effort to process and attend to distractor -let distractor go through late selection (pattern recognition), since left over mental capacity, -identity distractors -late selection in low load task has negative consequences when the distractor is competing - identify them - causing interference with primary task -High load condition: -competing distractor does not affect reaction time, response time for the compatible and competing condition is the same -When have high load, using all mental effort to process 6 circles that are all potential responses -have no mental capacity left to attend to distractor, cut off processing distractors ASAP, don't send through pattern recognition, ignore irrelevant stimuli --> early selection - - (ignoring them based on spatial location - never identify them as diamonds or squares --> hence they never interfere with the primary task) *shows people are sensitive to how much load capacity, naturally determine whether to use late or early selection

Selection and Task Load: -Video Game players

repeated their experiment using heavy video game players. -in both the low and high load conditions, the competing distractor slows them down (whereas in regular participants, only low load condition slowed them down) -for the video game players, the high load condition is really a low load condition -processing 6 circles at a time is no problem for heavy gamers --> There is still plenty of mental capacity left over to attend to the distractor appearing in the periphery. -have left over mental capacity to let distractors go through late selection --> causing interference with primary task, slowing reaction time -even at high load, have enough capacity to attend to distractor, -have more capacity than normal people -high load for video game player is like low load for non-video game player -argues that video game playing increases their mental capacity --> increase ability to multi-task -can train mental capacity and increase it

Examples of late selection: •Semantic Context Effects (Treisman, 1960)

showed that selection seems to occur only after we have assigned meaning to the stimulus. •involved dichotic listening task in which the participant was to shadow the left ear. •participant heard two different sentences in left and right ear. -1) I SAW THE GIRL/song was WISHING me that bird/JUMPING in the street -2) SITTING AT A MAHOGANY/three POSSIBILITIES let us look at these/TABLE with her head •#1 is shadowed/ attended ear, - #2 is unshadowed/ unattended ear. •upper case= what participant repeated back •notice that when a meaningful continuation of what had been played to the shadowed ear was played to the to-be-ignored ear ("jumping" is a meaningful continuation of "I saw the girl," but "song" is not), participants temporarily shadowed the to-be-ignored ear instead of the correct ear --> •Apparently, participants are basing their decision on what to shadow at least partly on the meaning of what they are hearing. → In other words, the selection of what to attend to and what to ignore is seems to be made AFTER the word is recognized, something that should not happen according to early selection theories. --> Pattern recognition BEFORE selection - (supporting late selection theory)

Dual Task Paradigm for Studying Capacity Models : Example - Johnson and Heinz (1972)

•As an example, Johnson and Heinz used the dual task paradigm to test the hypothesis that the higher up the processing chain we had to do selection, the more mental effort was consumed. Group 1: main task 1 + subsidiary task Group 2: main task 2 + subsidiary task •Dependent variable= performance on subsidiary task - Reaction time to light -Where performance on subsidiary task is worse, that version of the main task is hogging capacity -look at effects of using effort on main task on performance on subsidiary task -instructions to focus on main task first -groups differ in respect to what main task performed at the same time as the subsidiary task - differ in levels of mental effort •control group: performed subsidiary task only Group 0: Shadowed single list i.e. a single list of words was read to them and they simply had to repeat back the words. Although this task may take more mental effort than simply responding to the light, no real selection is necessary to perform the task —there is no requirement to discriminate one set of words from another. •Group 1: shadowed list in either male or female voice (hearing 2 simultaneous lists) -read two simultaneous lists, one by a female voice and one by a male voice, and they had to repeat back the words from either the male or female voice. - In this task, low-level selection is required. Female and male voices have to be distinguished from each other. Because some selection is required, more mental effort is presumed to be necessary in this task than in the Group 0's task. The selection, however, can be done on relatively low-level physical characteristics since male and female voices can generally be discriminated on the the basis of pitch. Certainly, there is no need to process the meaning of each word. Hence, the increase in required capacity should not be huge. •Group 2: Shadowed city names or names of occupations (hearing 2 simultaneous lists) *note how groups 1 and 2 differ in the level at which selection can occur. -also heard two simultaneously presented lists. They had to repeat back those words that were city names or occupation names. This version of the shadowing task requires higher-level selection. The meaning of the word must be processed in order to select which stimuli to attend to. Because the selection occurs higher up in the processing chain, more mental effort is required to perform it than the selection required by Group 1. Hence, Group 2 has less mental effort left over to perform the subsidiary task than does Group 1. Of the four groups, their performance on the light task should suffer the most. •Subsidiary task (all groups): press a button as quickly as possible to randomly presented light •The basic idea is that the more effort that must be spent on the main (shadowing) task, the less is left over to perform the subsidiary (light) task. Consequently, the more effort needed to perform the main task, the slower I should be to respond to the light.

Spotlight metaphor of attention

•Attention is like a spotlight focused on a particular spatial location -Brightest at center, weaker at fringes, dark at periphery •Space is medium of attention •What we pay attention to is a spatial location

Pre-attentive processes

•Automatic processes: -through training and consistent mapping, what was once an effortful process becomes effortless •Pre-attentive: -some processes are effortless because they are performed earlier in the system than when attentional capacity limits first occur -Do not require practice or learning -You can think of these as in-born processing capabilities Pre-attentive processes, like automatic processes, are effortless or near effortless. They are a bit different in the following sense. Whereas automatic processes once required effort but become automatic and effortless through practice and consistent mapping, pre-attentive processes are processes our cognitive system seems to have come equipped to be able to perform from the start without effort. -Subitizing, which we will discuss in a minute, may be one such process.

-Visual search demo - (another ex. of pre- attentive processing)- Visual search and Feature detection

•Because we have pre-attentive processes that detect these features, so if discriminate features are present, can find letter regardless of how many dissimilar letters are in background •When embedded among dissimilar targets (O among V's), the target is easy to find and reaction times do not increase as the number of distractors increases. •Looking for letter when background letters are not similar, doesn't matter how many letters - quantity of numbers doesn't matter •When embedded among similar targets, target is more difficult to find and reaction times increase (linearly) with the number of distractors. -The more distractors (b's) the longer the reaction time to detect the (R's - target) -when similar features

Dual task paradigm for studying Capacity models

•Capacity theorists have used the dual task paradigm to test their assumptions. •in the dual task paradigm, two groups of people perform the same subsidiary task, but different versions of some main task. •participants are told that performance on some main task is more important than performance on the subsidiary task, so they should firs make sure that they are attending to the task and then attend to the subsidiary task. •The dependent variable of interest is some measure of performance on the subsidiary task, sometimes accuracy, sometimes response time or latency (i.e., the length of time needed for a person to make a response to the stimulus) •worse performance on the subsidiary task in one group than in the other is interpreted to mean that the corresponding version of the main task requires more mental effort or capacity than the other version.

Early Selection: •Broadbent's Filter model

•Donald Broadbent formalized these thoughts into his Filter Model of attention, (illustrated in slide). • Began with the observation that we have multiple SENSORY CHANNELS feeding into us. Since most of his work used the dichotic listening task, he generally drew his model with 2 sensory channels, corresponding to the 2 different ears. - •However, in broader terms, a sensory channel can be anything that can discriminate two or more incoming stimuli based on low-level physical characteristics. - sensory channels may be based on left vs. right ear, pitch, spatial location of incoming sounds, intensity, etc. •he also believed that our pattern recognition processes -(those which allow us to identify what it is that we are seeing or hearing) - could only handle a limited amount of info. -Given inputs coming in on multiple channels, we need to cut some of those off before the info gets to the pattern recognition stage, otherwise, we would overwhelm it and not be able to attend to anything. -proposed that we have an ATTENTIONAL SWITCH to accomplish this goal. --> this switch shuts off all sensory channels but one. Only info from this one channels that has been left open gets through to the pattern recognition processes; protecting it from overload. -Broadbent's Filter model is an example of early selection - selection occurs BEFORE pattern recognition -told to shadow left ear (ignore right ear) attentional switch--> pay attention to 1,4,7, (while ignoring 3,2,5 from right ear) - can ignore other ear, or take physical characteristic and decide what channel going to let through based on that physical characteristic) -set the gate so only left sensory channel is being filtered to pattern recognition processes where 1,4,7 are identified

Task load and early vs. late selection

•Early selection on irrelevant stimuli reflects high-load tasks on relevant stimuli -primary task requires all available attentional capacity -there is no capacity left over to process other stimuli -therefore, they must be filtered out ASAP -ex.) Suppose that I am doing one task, and that it is a high-load task—it requires a lot of mental effort, like multiplying two-digit numbers in my head. Now another stimulus comes along. Since the first, high-load task consumes all of my mental resources, I cannot afford to let this stimulus get very far up the processing chain. I simply have no mental effort left over for it. So, if at all possible, I have to filter it out as soon as possible. That is, I'm going to have to do early selection. •Late selection on irrelevant stimuli reflects low-load tasks on relevant stimuli -primary task leaves some attentional capacity available -there is capacity to process other stimuli -therefore, they can be processed further up the chain -Ex.) But suppose that I'm doing a low-load task, one that does not require a lot of mental effort. Now I have capacity left over. A second stimulus comes along. Since my first task does not require much in the way of mental effort, I can afford to spend some effort on this second stimulus. I can process it further up the chain. I can perform late selection on it.

Attenuation vs. Late Selection

•Extremely difficult to experimentally distinguish the attenuation model from the late selection models (attenuation can predict the same effects that late selection would predict)

Examples of late selection: •Semantic Context Effects (Gray and Wedderburn)

•Gray and Wedderburn conducted a similar experiment. -The left ear hears "Dear 5 Jane." -The right hears "3 Aunt 4." •If asked to shadow the left ear, participants repeat back "Dear Aunt Jane." If asked to shadow the right ear, the participant repeats back "3 5 4." •Again, the decision as to what to repeat back is being determined by the meaning of the stimulus and hence must be occurring AFTER pattern recognition. -switching ears to get meaningful word. •The effect works all the way down to the syllable level. If the left ear hears "OB 2 TIVE" and the right "6 JEC 9," and the participants are asked to shadow the left ear, they will construct a meaningful whole from both inputs and repeat back the full word "objective."

Capacity models: An analogy

•I can lift 100 lbs •I can divide that capacity across many different objects -100 lbs of grapefruit -100 lbs of oranges -75 lbs of oranges plus -25 pounds of grapefruit -But NOT 60 lbs of oranges and 60 lbs of grapefruit •We can make a fairly direct analogy between mental effort and physical strength. •Suppose that you are able to lift 100 pounds. You can choose to use that strength to lift any combination of things so long as the total weight is not more than 100 pounds. If it is more than 100 pounds, then we are going to run into some kind of difficulty or another. •The same applies to mental processing. We have so much strength. We can choose how to divide it up amongst various mental tasks and so long as all the things that we are doing do not require more mental effort than we have available, we are okay. • As soon as they do exceed the amount available, though, we start making errors, slowing down, and stopping some mental tasks.

Determining the primitive or basic features of the perceptual system: -Pop-out method

•If two areas have different features, or different values of feature, then a boundary between the two areas immediately pops-out •Pop-out has been proposed as a way of determining what the basic or primitive features of the perceptual system are. •If objects with one value of a feature pop-out against a background of other objects with a different value of the same feature, then that feature is a primitive feature, - i.e., a feature that the perceptual system uses to identify more complex objects. •(We will be discussing this a bit more when we talk about object identification.) In this example, both line orientation and line crossing would be considered a feature. "Slope changes," however, would not be considered a feature.

Treisman's Attenuation Model

•In her attenuation model, Treisman argued, in keeping with early selection models, that selection does in general take place before pattern recognition. --> •However, the filter does not completely block out unattended channels. Instead, it merely ATTENUATES (to turn down) them, or reduces the strength of the information coming through the filter on the unattended channels. ---> Filtering does not completely turn off the unattended channel. It simply turns down the volume. That's the key assumption. -unattended stimuli input coming in at much reduced volume (vision coming in dimly, noise quietly) •Since Treisman did most of her early work with the shadowing task, we'll discuss her model in terms of recognizing spoken words. •In her model, each word or concept is represented by a node in a mental lexicon or mental dictionary. •Each word has a threshold "subjective loudness" which must be exceeded for that word to be recognized. -how loud have to hear word in order to recognize it, varies by subjective importance to person (i.e. words like your name, to recognize have relatively low thresholds, can hear at low volume and understand it, - but words uncommon to you, or low significance have high thresholds --> must hear at high volume to recognize) •In keeping with the assumption that filtering occurs before pattern recognition, the subjective loudness corresponds to the strength of the signal coming through the filter. •Words on the shadowed channel come through loud and clear; those on the unattended channel come through at a reduced loudness, as indicated by the upward pointing arrows towards on the left and right sides of the figure. - The height of the arrow indicates the loudness of the message. •All words on the attended channel get recognized, because the signal comes through with adequate strength. - But only words with LOW thresholds can be recognized on the unattended channel. -Words with high thresholds will not be recognized. --> (subjective volume - words that have low threshold in order to recognize it) •What determines a word's threshold? Two things. •First, some words are just always important to me—my name, my wife's name, the name of my sailboat. Important words have permanently low thresholds and can thus nearly always be recognized (hence, the cocktail party phenomenon). •Second, the threshold of words is temporarily lowered when they become relevant due to the context. - Contextually important word --> temporary lower threshold when word is in text. - Hearing "The girl was" temporarily lowers the threshold of words such as, "jumping, singing, running, asking," etc., etc., the whole list of verbs that girls can generally be doing. --> A relevant word can come through the filter with very reduced volume but still be recognized because its threshold for recognition has been temporarily lowered. •Hence the results of the Treisman 1960 experiment - city names presumably have a temporarily lowered threshold when in the context of an experiment they predict shock, -even if normally have high threshold, temporarily lower threshold in non-shadowed ear bc important in context. Hence the results of the Corteen and Wood study.

What is learned when a process becomes automatic?

•Increased speed approach -With increased practice and consistent mapping, processes get faster and faster -With sufficient practice, speed becomes so quick that becomes instantaneous •Instance-based view -With practice, store in brain, association between stimulus and response made -Example, every time see light → consult long term memory and retrieve from memory what to do •When new driver, don't have many instances in brain to break, → so takes longer •When good driver, have many instances in brain that associate red light with breaking, → quick response, → automatic - A fair question to ask concerns what changes when a process becomes automatic. This question has been answered in two different ways in the literature. (1) The increased speed approach argues that the component processes making up the overall process all get faster. In addition to getting faster, some of those sub-processes may transform from serial processing to parallel processing. -Serial processes are performed one after the other; -parallel processes can all be performed simultaneously, thus saving time. (2) An alternative view is the instance-based view. According to this view, as a process becomes automatic, the processing changes in a fundamental way. Rather than actually carrying out the process, we retrieve an earlier instance of having done so along with the result of that processing. We then base our response on having retrieved the result of that processing. -Note that the increased-speed view and the instance-based view are not mutually exclusive. One may apply in some situations (Perhaps the increased speed approach applies to braking for a red light) and the other in other situations (Perhaps the instance-based view applies to doing mental arithmetic.).

Treisman's Attenuation Model (cont.)

•Mental lexicon with variable thresholds -each word or concept represented by a node in a mental lexicon -when a node's activation level exceeds its threshold, we become conscious of that word -thresholds vary across words based on their subjective importance (how loud must hear word in order to hear it) -important word =low threshold -thresholds vary from moment to moment, based on context -unattended channels are attenuated, not completely unheard -Therefore, with context, enough meaning leaks through to drive a concept or word with temporarily lowered threshold (or a concept or word with a permanently low threshold) above its threshold --> recognize the word,

Visual Neglect

•Neurological disorder resulting from brain damage •Patients fail to respond to objects contra-lateral to a brain lesion •Interpreted as a failure to distribute attention to that side of space

Cueing Attention

•On each trial, a single letter is presented- -½ trials, stimulus presented to the left of fixation point -½ trials, stimulus presented to the right of fixation point •Task: press the left or right button as quickly as possible, depending upon which side the letter (stimulus) appears •Cueing: Immediately before presenting letter, one side or the other is cued (arrow=cue) -valid cue (.8): Side on which target actually occurs is cued • cue pointed toward same direction that letter is presented • cue points to right and letter is on the right -Invalid cue (.2): side opposite to that on which target occurs is cued •Cue pointed toward different direction than letter is presented •Cue points to the right and letter is on the left •General result: RT is faster on valid cued trials than invalid cue trials (or non-cued trials) • Neutral= cue points to no direction • Shows people can direct attention to location in space •If cue tells you on right, can still stay fixated straight ahead, but attention directed to the right (can attend to a location in space)

A shocking failure of Early Selection •Corteen and Wood (1972)

•Phase 1 of experiment -participants listened to a list of words •city names followed by an electric shock --> GSR (measure nervousness) •Other nouns •Phase 2: Dichotic listening task -unattended ear: •3 old city names •3 new city names •3 old nouns •3 new nouns -participants could not report what they heard on unattended ear -But, GSR higher to city names (both old and new) --> •the condition was present, thus processing meaning of that stimulus, processing meaning that it was a city name because showed through GSR level - conditioned response --> processing meaning on unattended channel, pattern recognition BEFORE selection •This example is important because it shows that even when we claim not to be able to report anything about the non-attended stimulus, we still may have processed information about its meaning at a subconscious level. •In the first phase of their experiment, Corteen and Wood had people listen to a list of nouns, some of which were city names. Whenever a city name was read, the person was shocked. No shock was delivered for non-city names. --> Soon, people were conditioned to expect a shock when a city name was read. This could be determined by measuring their galvanic skin response or GSR. •The GSR is a measure of the electrical resistance in the palm of the hand. When we sweat, our palms become wet and the GSR goes up. When we get nervous, we sweat. (This is the primary response used in lie detection tests. Presumably, when we lie, we get nervous, so we sweat, so our GSR goes up.) Soon, listeners in Corteen and Wood's experiment began to sweat, in anticipation of the shock, whenever a city name was read. (The sweating began even before the shock was actually delivered.) •The second phase of their experiment involved a dichotic listening task. •To the unattended, to-be-ignored ear, words were read. No shocks were delivered in this phase of the experiment. Some were city names from the first part of the experiment; some were new city names, not used in the first part; some were non-city nouns used in the first part of the experiment; some were new non-city nouns. •Listeners were unable to report back which old city names and nouns were read to the to-be-ignored ear. Similarly, they were unable to report any of the new city names or new nouns. •Nevertheless, they showed an elevated GSR to the city names read to the unattended ear (both old and new), apparently anticipating a shock. --> • So, even though they could not report back these stimuli, they had processed them far enough along at a subconscious level to have identified them as city names. •Again, we see that attention (thus, selection) seems to not kick in until AFTER pattern recognition, not before, a result inconsistent with early selection.

How does a process become automatic?

•Practice -Become good at reading by reading •Consistent mapping between stimulus and response -We always make the same response to a given stimulus •Ex.) We have an automatic response of braking when we see a red light -We would not if in different cities different colored lights were used for "stop" -Two things are necessary for a process to become automatic. (1) First, the behavior must be well practiced. We read words automatically because we are so highly practiced at doing so. Many parts of driving are automatic because we do them all the time. (2) Second, the mapping between the stimulus and the response that we automatically make to that stimulus must be consistent; the response to the stimulus must be the same in all contexts. -We would never learn to stop (the response) automatically when we see a red traffic light (the stimulus) if in different cities or different states, red sometimes meant to stop and sometimes meant to go. -And we rarely encounter groups of letters where we are not supposed to read the word that they represent. -Note that practice and consistent mapping are both necessary in order for a process to become automatic. Given just one without the other, the process will not become automatic.

Cost of inhibiting an automatic process - -Besner, Stolz, and Boutilier RESULTS

•Results → 2 effects: (1) Reduction of stroop effect in letter condition, (less difference in reaction time between incongruent and congruent trials in letter condition than in congruent and incongruent trials in word condition) •reduction of congruency affect in letter condition → suggest that we can inhibit automatic processing (2) However, overall response time were longer in letter condition → shows paid price for inhibiting automatic processing, → longer overall response times in letter condition than word condition

Examples of tasks performed automatically

•Riding a bicycle (to a certain extent) •Driving a car (to a certain extent) -Yet, if started sliding on ice, no longer automatic but controlled processing used •Greeting a person •Reading print •Returning a tennis serve? •Hitting a fast ball? → these two perhaps some components of automatic processing, but require controlled processing, active/conscious resources used well. If skilled tennis player, some processed are automatic, but if beginner, require more controlled processing -Riding a bicycle and driving a car are qualified as being automatic "to a certain extent" because although they are mostly under automatic control, when an emergency occurs, we may need to break out of the automatic mode and revert to more controlled processing. Such would occur when, for instance, a child runs out in front of the car we are driving, or when a car turns directly towards the bicycle that we are riding. •Many aspects of skilled athletic performance are also likely to be automatic, at least to a degree. Skilled athletes seem to execute their skills without really thinking about them. Some even say that thinking too much about what they are doing interferes with their performance. They do not seem open to conscious awareness. •However, these skills cannot be completely automatic; they do require concentration. Athletes frequently speak of the need to focus in order to perform well. Certainly, it would be very difficult to return a tennis serve while at the same time solving algebraic equations in our head. Hence the question marks by the last two items.

Early vs. late models of selection •Late models of selection

•Sensory store→ Pattern recognition→ Selection →Short-term memory • Late selection argues that filtering is not necessary because pattern recognition is NOT capacity limited (all hinges on one's view of pattern recognition) --> consequently, the processing stage we had labeled Filter is not necessary. The first capacity limitation occurs at short-term memory, so we can afford to delay dropping stimuli until AFTER pattern recognition, at the Selection stage. -- so can identity everything and then decide what going to attend to and select for short and long term memory --can use meaning of stimulus (e.g. recognize name is spoken) to decide for selection what going to attend to and what going to filter out

Early vs. late models of selection - •Early models of selection

•Sensory store→ filter →pattern recognition→ Short-term memory •According to early selection theorists, the stage we had labeled selection is not necessary, bc we have filtered out all the information we need to BEFORE the pattern recognition stage. (in which just have the physical characteristics of stimuli, e.g. sound, location, vision -color, etc. -no words) --> •The reason we do that is because pattern recognition is capacity limited and we need to drop info before pattern recognition in order to avoid overwhelming it with stimuli and causing a general breakdown in processing

Examples of Early selection: Experimental Paradigms •Dichotic listening task -Shadowing

•Shadowing (Cherry, 1953; Moray, 1960) -Dichotic listening task: Separate messages played to left and right ear •participants told to repeat back message spoken to either the left or right ear -later, recalled practically nothing of the unattended message (non-shadowed ear) --> Did detect changes in gender of speaker and other low-level physical characteristics •led to the idea of filtering BEFORE pattern recognition (before determine meaning of stimulus) •A basic finding from this literature is that when, say, the left ear was shadowed, people could remember nothing that was said to the right ear (unattended/non-shadowed ear). Changes in the physical characteristics of the message to the right ear (unattended ear), however, were noticed - I.e. changes in gender (i.e. pitch), changes in loudness --> •these results were generally interpreted to mean that nothing about meaning gets through on the unattended channel, but only low level physical characteristics of the stimulus --> (supporting early selection theory of filtering out stimuli BEFORE pattern recognition)

Locus of Attention

•We can attend to either a location in space or to a particular object (that is possibly moving through space) -Can be toward particular object (point a) •Probably can be strategically influenced -i.e. the choice is ours •we do not necessarily attend to a whole object, but to part of it -when attend to particular part of object, attention tends to spread within object than across objects • faster to attend to B then C when attention is focused on A → attention is spreading more strongly within object A is in then across objects -attention moves through space more easily within an object than across objects

Controlling Attention

•We have some control (perhaps unconscious) in the ability to perform late selection or early selection •What determines our "choice"? -Task load (i.e. amount of capacity being used) - amount of mental effort we are currently expending on other things if: •High load (under a lot of mental effort) -need to shed extraneous information quickly •Early selection - (to stop attending to extraneous info and focus on important) •Low load -can carry extraneous information to higher levels of processing •late selection (- might as well process extraneous info, to the level of meaning or beyond - have the capacity for it, can afford to do late processing)

Locus of attention

•What is it that we can attend to (in the sense of focusing our attention)? -A spatial location? -A particular object? -Our choice?

Late Selection

•When attention capacity exceeded, unattended events are recognized - but - -not selected for access into short term memory -quickly forgotten about because they are not rehearsed. •context causes already recognized pattern to be selected for entry into short term memory •The second response to experimental findings that meaning can in fact determine what is attended to was the development of late selection models. •These models assume that the pattern recognition stage is not capacity limited—it can handle whatever amount of data we wish to throw at it at any given time. •What is capacity limited is short-term memory; we can only hold so much information in short-term memory at once. •Consequently, we can let everything go through pattern recognition, but we then need to throttle things back before letting things go to short-term memory. --> The Selection stage is responsible for determining what information gets into short-term memory and what information gets dropped. •Information that is dropped is not well remembered, since it never gets into short-term memory where it can be rehearsed and transferred to long-term memory. •Furthermore, because Selection operates AFTER pattern recognition, we can use the output of pattern recognition—i.e., our identification of the stimulus—to decide whether to drop or attend to that stimulus. •That is where context comes in—if a word (in a listening experiment) is relevant because of the context, since we have already identified the word, we can now decide to attend to it based on it being relevant to the context and put it into short-term memory for additional processing. •The key assumption in these models is that selection occurs AFTER pattern recognition.

Visual search and feature detection

•When target and distractors do not share features -detecting any single feature is sufficient for detecting the target • speeds search time • Number of distractors does not affect search time • pop-out •When target and distractors do share features -Distractors need to be inspected one by one to check whether they are the target • Slows search time • Increasing number of distractors increases search time • Don't pop out •But, how do we know what qualifies as a feature?

Control of Attention: •What determines what we attend to? -Endogenous control

•or Goal directed Selection: =refers to our ability to choose what to attend to. since we presumably make this choice based on whatever it is we are trying to accomplish at the moment, this form is control is a.k.a. goal directed selection. *keep in mind that the word "choose" is not meant to imply that we make a conscious, deliberate choice as to what to focus on. The choice may be quite subconscious -we determine how our attention is allocated -Not necessarily a (phenomenal) conscious action •Endogenous= internal

Control of Attention: •What determines what we attend to? -Exogenous control

•or Stimulus-driven capture: =refers to the seemingly automatic capturing of attention caused by certain external stimuli that we have no control of. e.g.) -hearing your name -appearance of a new objet -loud noise -motion •very much reflexive --> stimulus driven •Exogenous=external

Key claims of Early Selection Models

•pattern recognition is capacity limited •info comes to us on a variety of different physical channels -we can attend to only one channel at a time •attention operates at a physical, sensory level -we select what channel comes through based on some physical characteristic of the input •Attention is necessary for deriving meaning -filtering occurs prior to pattern recognition/object identification -therefore, meaning cannot be used to decide what to attend to -in order to identify what a stimulus is - what object is depicted by a particular pattern of stimulation on our retina, or what word a sound waveform corresponds to - we need to first apply attention, so that that stimulus and only that stimulus gets into our pattern recognition devices. - Otherwise we would overwhelm those devices and end up not recognizing anything. only by applying attention are we able to subsequently derive meaning -and since attention is necessary for deriving meaning, we cannot use meaning (i.e. a stimulus's identity) to decide what to attend to. We have to use other physical characteristics instead.

Allocating our attention (Endogenous Control): •Early selection -General information processing model

•refers to filtering information out before we even identify it, - the filter stage -the basic idea is that there is so much typically happening in our immediate environment that we cannot possibly identify and process it all. --> we must do something in order to make the amount of information that we are dealing with manageable. otherwise, the pattern recognition processes get overwhelmed and we would be confused. -so we filter our stimuli as early as possible in the processing stream, BEFORE the pattern recognition stage --> because we are filtering out stimuli before the pattern recognition stage, we can't use their identity, or their meaning, to decide what to drop and what to keep - we are deciding what to filter our before we even know what the stimulus is. - all we can rely on is the physical characteristics of the stimuli to decide what to filter out and what to keep. -E.g. For visual stimuli, these would include characteristics such as color, brightness, and spatial location. -for Auditory stimuli, characteristics would include pitch, loudness, and spatial location.

Late Selection Models of Attention: -key claims

•significant claim: Many "attentional deficits" are not because the perceptual system cannot keep up, but because short-term memory cannot keep up -everything is perceived/identified, but may be forgotten before we become conscious of it, or are able to process it such that it can make it to Long term memory -Short term memory has a limited capacity -attending to one item in short term memory causes loss of other items •What late selection models are saying is that we perceive and identify everything that is going on around us, at least at a subconscious level. •Pattern recognition is not a bottleneck. Short-term memory, though, is—it can only hold so much information and so we need to be judicious about what information we let into short-term memory. -(Many people equate what we are conscious of with the information in short-term memory, so for these people late selection models are basically saying that we can only be simultaneously conscious of a limited number of things.) -pattern recognition has an unlimited capacity, thus no filter needed before; -short term memory is capacity limited --> so filter in selection stage AFTER pattern recognition, - to determine what info goes to short term memory (what to ignore and what to store)

Is there perhaps some truth to both early and late selection? -can we predict under which circumstances one or the other will occur?

•sometimes we do early selection and sometimes we do late selection --> not mutually exclusive (understanding attention requires understanding both) •Early vs. late (and attenuation) selection has been debated for a long time within cognitive psychology, some findings supporting one approach and other findings supporting the other approach. •This has led some psychologists to suggest that perhaps we can do both early and late selection, that the choice is somehow up to us. • It's not enough, though, simply to say we can do both. •For such an approach to have any substance, we must be able to say under what circumstances we do early selection and under what circumstance we do late selection. And we can do that. In order to understand our answer, we need to understand what have been termed capacity models of attention

•Late selection -General info processing model

•those on the late selection side of the debate argue that the pattern recognition mechanisms have no capacity limitations. -we can throw however much info •The first time we encounter limitations in terms of how much information we can handle is in short-term memory, which is very limited in terms of how much info it can handle. •in order not to overwhelm short term memory, we need to be careful concerning what info we decide to allow in there. --> we do that through the SELECTION stage of processing. •because pattern recognition is not capacity limited, can perform the selection AFTER pattern recognition occurs. --> this means that we can use a stimulus's identity or its meaning to make that selection of what proceeds or doesn't proceed to short-term memory •The ability to use meaning to decide what stimuli to focus on is a chief difference between early selection theories (which say we cannot use meaning) and late selection theories (which say we can use meaning)