4. Semantic Face info

Question 1

what do we need to do once a face is recognised?

Answer 1

Once a face is recognised, we typically need to retrieve what we know about the person (semantic information) who they are, what they do.

Question 2

what two types of semantic info is there?

Answer 2

info:
- Visually-derived semantic information: information seen or inferred without needing to recognize a face (e.g., ethnic group, sex, age)

• Identity-specific semantic information: can be accessed for recognized faces (e.g., occupation, place of living, hobbies etc.), generally not revealed by visual characteristics. You only have this information for people that you know - FOCUS ON THIS ONE

Question 3

what is face recognition/ person recognition assumed to be?

Answer 3

Recognising people as accessing identity-specific semantic information is often seen as a multi-stage hierarchical process involving sequential access to different kinds of information (Bruce and Young, 1986).

Question 4

what three hierarchical stages is face recognition broken into?

Answer 4

1. Recognising a face as familiar
2. Accessing stored semantic information
3. Retrieving a person’s name

Question 5

what do the three hierarchical stages form?

Answer 5

A sequence of this kind formed the main identification route in the Bruce and Young (1986) model of face recognition.

Question 6

early study to support the 3 step sequence what did they do?

Answer 6

One of earliest studies that supported this sequence was the- Young, Hay, & Ellis (1985)

• Diary study (people carry diary and have to make an entry whenever they realise that something happens): 22 people took notes/ description of what happened on mistakes in recognising people across an 8-week period.

Question 7

what did Young, Hay and Ellis (1985) find?

Answer 7

• Records grouped into three categories, reflecting errors at different levels of recognition:
• Failures to recognise a familiar person, mistakenly thought to be unfamiliar.
• Recognising a person as familiar, but being unable to bring to mind other details (e.g., occupation and name).
• Recognising a person as familiar and remembering semantic information, but failing to remember the name.

Question 8

what can we infer from the Young Hay and Ellis study?

Answer 8

Because there are these three errors and some errors didn’t exist e.g. people never could access the name of someone but know nothing else about the person (e.g. what they study) because some types of errors but not others people also thought that there existed three serial stages 1) recognition as a familiar face –> 2) stored semantic information about an identity –> 3) name retrieval.

Question 9

what are problems with Young, Hay and Ellis’ Study

Answer 9

However, diary studies taken alone have a number of limitations (Reason & Lucas, 1984). The most important of these are that diarist’s reports might show biases, perhaps because they only report the errors they find particularly striking or easy to interpret themselves, or because they deliberately or unintentionally distort the errors in ways that make them seem more clear-cut than was actually the case.

Question 10

what has additional confirmations of findings of diary studies thus been obtained from?

Answer 10

Additional confirmations of the findings of diary studies of everyday errors has therefore been obtained from studies of errors made under lab conditions. These remove reporting biases because errors can be systematically examined as they are made, yet still find the types of error predicted by the sequential access view.

Question 11

further evidence for sequential processing

Answer 11

Hay et al., (1991)- showed 190 photos of famous or unfamiliar faces to 12 participants who were asked whether or not each face was a familiar person, what the person’s occupation was, and what the person’s name was. All of the errors made fell into patterns which would be expected from the hierarchical sequential access model, and examples were found for all of the types of error predicted by such a model. No types of error that would be incompatible with a sequential access model.

Question 12

further evidence for sequential processing (study on giving cues)

Answer 12

Hanley and Cowell (1988)
Participants who found a face they should have recognised to be unfamiliar, familiar only, or who knew who the person was but couldn’t remember the name, were cued by giving them semantic information about the person (e.g. brilliant golf player) or a card containing the initials of four famous people (one being the person in question), with blank spaces for the remaining letters. The semantic cue was most effective at promoting correct naming when participants found the face familiar only, whereas cueing with the initials was more effective when the occupation was already known. ]
This is exactly as would be expected from a sequential access model, since people who find the face familiar only would be ‘blocked’ at the stage where semantic information would normally be retrieved, whereas people who can already access the occupation but are still searching for the name should derive more benefit from an initials cue than a semantic cue.

Question 13

aside from experimental experimental studies what evidence is there for sequential processing?

Answer 13

Neuropsychological impairment as evidence

Question 14

what cases of Neuropsychological impairment are there?

Answer 14

• prosopagnosia- known faces unfamiliar
• can have cases where faces are recognised but access to semantic info or names can’t be accessed
• anomia- name retrieval to known faces is problematic even though semantic information can be properly accessed.

Question 15

what other evidence exists for sequential processing

Answer 15

Reaction time experiments for support of the sequential model

Question 16

reaction time to names and semantic info

Answer 16

Faces can be classified as familiar fare more quickly than they can be classified by occupation and categorisations based on occupations or other semantic properties can usually be achieved more quickly than categorisations which require access to the person’s name.

Such findings hold even when task demands and response requirements are carefully equated (Young et al., 1986; Young et al., 1988).

Question 17

study on slower access to name outputs and what did they do?

Answer 17

Slower access to information about name outputs can arise even when the ‘semantic’ information needed seems intuitively more demanding. E.g. Johnston and Bruce, 1990:

• 8 faces used repeated throughout the experiment
• 4 called john and 4 called james
• Half British and half American celebs
• Half dead and half alive
• Participants shown pairs of faces and asked to verify as quickly as possible whether the two faces matched or mismatched on a dimension specified separately for each block of trials. In one block of trials the faces were to be matched by name, on another block they were to be matched on nationality and on a further block they were to be matched in terms of whether both faces were dead or both alive.

Question 18

what did Johnston and Bruce find?

Answer 18

• Found that all the tasks involving matching for names were conducted more slowly than any of those assessing nationality or dead or alive, even though intuitively it seems that it should be harder to retrieve information about whether a face belongs to a person who is now dead than whether they are called john.

Question 19

what does the Johnston and Bruce study suggest?

Answer 19

Again support three-stage model of person identification- names must take longer as name retrieval requires an additional and time consuming stage.

Question 20

what did findings on sequential processing inform?

Answer 20

The findings above about the stages of retrieval of personal identities from face inputs formed the core of the face identification route in the Bruce and Young Model (1986)

Question 21

Outline the Bruce and Young (1986) Model?

Answer 21

• Personal identity is accessed for a familiar face via activation of “Face Recognition Units” (FRU).
• Appearance of each known face is stored in an FRU.
• FRUs become activated when a “structural description” (visual input) matches a “stored description”.
• “Person Identity Nodes” (PINs) hold semantic information.
• PINs can be accessed from different input domains (e.g., hearing a voice, seeing a printed name).
• Names represented at a separate stage, in line with the sequential access idea.
  This model basically shows this same order. With a face recognition then semantic info then name stage.

Question 22

According to the Bruce and Young Model what happens when faces are viewed?

Answer 22

According to this model, when a face is viewed, a number of different kinds of analysis are conducted in parallel. These include the derivation of information about emotional states (via analysis of facial expressions), and information about visual speech (from lip, tongue and jaw movements), as well as personal identity (identity-specific and name codes).

Question 23

how is personal identity accessed in the Bruce and Young Model?

Answer 23

Importantly, though, personal identity can only be accessed for a familiar face via activation of the appropriate ‘Face Recognition Unit’ (the FRU, using the idea that the appearance of each known face is stored as a distinct FRU) when a suitable structural description of a viewed face matches the stored description for a known face

Question 24

what will/ will not a FRU respond to in the Bruce and Young model

Answer 24

A face recognition unit will respond when any view of the appropriate person’s face is seen but will not respond at all to his or her voice or name The person identity node, in contrast, can be accessed via the face, the voice, the (written) name or even a particular piece of clothing (only access via the face is shown).

Question 25

what happened when the Bruce and Young Model was developed?

Answer 25

When people developed these ordered models they also thought about how to test it… tested with Repetition and Semantic Priming

Question 26

what was the Bruce and Young Model tested with?

Answer 26

Repetition and Semantic Priming

Question 27

what is repetition priming

Answer 27

• Repetition priming: facilitation of recognition of a previously seen item (e.g., faces, words or objects). If you see something twice when you see it for the second time it will be easier to do something with it ie. with a face for the second time it will be easier to recognise.

Question 28

what are the two different paradigms of repetition priming?

Answer 28

• Immediate repetition priming

- Long-lag repetition priming:

Question 29

what is immediate repetition priming?

Answer 29

• Immediate repetition priming: Prime and target (50% familiar, 50% unfamiliar) follow each others immediately (as in N250r studies …)
• Task: familiarity decision on target- distinctly a prime and a target directly after. And have a repetition condition and non-repetition condition. People faster if seen same face immediately before.

Question 30

what is long-lag repetition priming?

Answer 30

• Primes and test items (50% familiar, 50% unfamiliar)- present multiple faces and have a test face in which some of the faces are familiar and some are new. So is not immediately presented as you will have to look at other stimuli before making a judgement. Task is again to say if the face is familiar. Speeded familiarity decision at test.

Question 31

what have priming effects provided?

Answer 31

Priming effects have provided important insights into the organisation of mechanisms involved in the recognition of familiar people. In particular, comparisons of repetition and semantic priming have been instructive.

Question 32

where are effects of priming biased to?

Answer 32

The effects of repetition seem to be located in the face recognition system itself, since decisions about the face’s expression or sex, which can be made without needing to recognise the person, do not show repetition priming ( Ellis et al.,1990)
pattern points to the locus of the benefit of repetition priming being in the system that recognises the face’s identity. When we look at a face, recognising it is involuntary – it is something we do even when we are focused mainly on some other aspect such as expression

Question 33

study on basic effect of priming?

Answer 33

Familiarity decision on familiar face faster if the face appeared previously in the experiment (Bruce & Valentine, 1985).

Question 34

what else did Bruce and Valentine, 1985 find about priming?

Answer 34

• Repetition priming of familiar faces is domain-specific
• Recognition of a face NOT facilitated by previously having read the name. means you can present a face and only a face.
• Effects show substantial image-dependence.
• Reduced, but still significant priming if test image differs from image of the same person studied earlier.
• Priming effect is relatively long-lasting.
• Survives presentation of intervening items, lasts for several hours (Bruce et al., 1998).

Question 35

what seems to be the only thing that matters in priming and study for this?

Answer 35

Doesn’t really matter what happens during priming as long as people are looking at the faces…evidence for this….

• Ellis et al. (1990): does it matter what task people have during priming? when you have a task that explicitly means that they have to process identity. When you ask them if it is a male or a female face for example, you don’t ask them anything about familiarity so they don’t have to process this but do people do this automatically?
• During initial presentation (priming phase), participants judged faces either according to sex, expression, or familiarity.
• Task at test: familiarity decision
• Repetition priming effect at test did not depend on task during learning.
• Conclusion: Face recognition is involuntary and to some extent automatic- we cant ignore identity of a face.
• We recognize a face when we are focused on other aspects of the stimulus.

Question 36

evidence that repetition priming effects can be remarkably robust?

Answer 36

when we are focused mainly on some other aspect such as expression.
Repetition priming effects can be remarkably robust. Bruce and colleagues placed faces on participant recruitment posters in university halls of residence, inviting people who could recognise the faces shown to come into the laboratory to take part in face recognition experiments (Bruce, Carson, Burton, & Kelly, 1998). Unknown to the participants some of the items that had been on the recruitment posters reappeared in a face familiarity decision task in the same or different pictures, and familiarity decisions to these items were made significantly faster than to items that were novel in the experimental lists. Again priming effects were greater to faces shown in identical pictures compared with ones shown in changed pictures. These effects showed not only that priming effects can cross a change in context (from the recruitment poster to the laboratory), but also that they can last for several hours or days, between seeing the poster and participating in the advertised experiment.

Question 37

explaining repetition priming with Bruce and Young model

Answer 37

• Is it possible to explain using Bruce & Young (1986) model?  explains repetition priming by assuming that activation of FRU (face recognition units) is still above baseline when face is presented for the second time. If we see a face for the first time faces activate FRUs and to recognise it it needs to exceed a certain threshold, this slowly decays once the face is no longer presented. If the face is then presented again then this will help because the same FRU will have some residual activity so the threshold will be reached more quickly.

Question 38

what is the other from of priming?

Answer 38

• Semantic (or associative) priming

Question 39

what is semantic (associative) priming?

Answer 39

Semantic priming tasks investigate the effect of having previously recognised a closely associated stimulus – for example, the effect of having recently recognised one member of a due (e.g. brad pit and Angelina Jolie)

• Semantic (or associative) priming: presentation of a closely associated stimulus facilitates recognition of a subsequent face.

Question 40

original study on semantic priming

Answer 40

• Bruce & Valentine (1986):

Question 41

what did the original study on semantic priming do?

Answer 41

• Bruce & Valentine (1986):
• Presentation of pairs of familiar and unfamiliar faces (similar to immediate repetition priming)
• Task: familiarity decision on the second face (target).
• Familiar targets preceded by (1) related (Brad Pitt  Angelina Jolie), (2) neutral (unfamiliar face  A.J.), (3) or unrelated (David Cameron  A.J.) prime faces.
• Varying time between prime and target (stimulus onset asynchrony, SOA).

Question 42

results of the original study on semantic priming

Answer 42

• Results
• Recognition facilitated by related prime stimulus onset asynchrony
• No evident cost for unrelated relative (familiar but no relation) to neutral primes (unfamiliar faces).
• Priming at a short SOA (250 ms) and absence of inhibition in unrelated (relative to neutral) trials suggests priming is not based on consciously anticipating specific target (expectancy).

Question 43

what do Bruce and Valentines (1986) results on semantic priming suggest?

Answer 43

• Presence of strong priming and absence of inhibition in unrelated trials suggests that the mechanism for these priming effects is highly automatic rather than based on participant strategies.

Question 44

what could be a possible problem with the Bruce and Valentine (1986) study?

Answer 44

people could be making predictions on what the next prime will be and that in many times they will be right- but this is thus not what we are interested in as seems to be more about strategic memory for individuals
Could be an active process

Question 45

what experimental technique can be used to see if active processes are behind semantic priming

Answer 45

• timing

- inhibition effect

Question 46

talk about timing and showing active processes arent behind semantic priming

Answer 46

Bruce and Valentine (1986)

Showing that presentation at very quick SOA (250) ms effect still exists

Question 47

talk about inhibition effects

Answer 47

if people make a prediction for what is next and then someone else occurs that this is a disadvantage- because people are having to move from their predictions to what is actually there.
So in theory people should actually be slower in the unrelated prime condition compared to the neutral condition, because in neutral condition there is no prediction that can be made.
This is called an inhibition effect

Question 48

What can be said for inhibition effects in Bruce and Valentine (1986) study?

Answer 48

They do not find them thus what we see is not a strategic process and also it is working at a very quick rate.

Question 49

So overall what can be said about repetition and semantic priming

Answer 49

There is some kind of activation flowing through some kind of semantic network and semantic priming works on automatic spreading of this information

Question 50

What are differences between long-lag repetition priming and semantic priming

Answer 50

Semantic priming produces clear benefits on recognition that differ from repetition priming effects in two important ways:

1. Facilitation produced by semantic priming is very short-lived.
2. semantic crosses stimulus domains

Question 51

evidence that Facilitation produced by semantic priming is very short-lived.

Answer 51

• Bruce (1986) varied the number of items that intervened between ‘prime’ and ‘target’ items in a continuous series of faces in which some items were repeated and others preceded by associates. Repetition priming effects of repeated items showed no sign of reduction as the number of intervening items varied from none (immediate repeat) to 11. Semantic priming effects were found only when prime and target followed each other immediately. Either semantic/associative priming lasts only a matter of seconds, or it is obliterated by an intervening, unassociated item.

Question 52

evidence that semantic priming crosses stimulus domains

Answer 52

• E.g. from the recognition of Brad Pitts face to recognition of Angelina Jolie’s name (Bruce and Valentine, 1986)

• Young, Hellawell, & de Haan, 1988: Unlike long-lag repetition priming, semantic priming crosses stimulus domains.
• E.g., recognition of famous face facilitated if directly preceded by written name of highly associated person, e.g.,
• face immediately preceded by name has effect
• can also go from faces to names with this effect

Question 53

what can be said about semantic and long-lag repetition priming?

Answer 53

seems the two effects are quite difference

Question 54

what do differences in semantic and repetition priming imply?

Answer 54

These differences imply that the sources of facilitation must arise at different loci in the recognition system. – Consistent with this ERP studies show that repetition priming affects a relatively early and topographically distinct N250r (Sommer et al., 1995). Semantic effects are in contrast more apparent in a later component of the ERP waveform the N400.

Question 55

what is the issue of semantic priming for the Bruce and Young model (1986)

Answer 55

The Bruce and Young model (1986) makes these effects had to explain.

Effects have different loci,

from the experiments we might think that the mechanisms behind these two effects are very different yet in the Bruce and Young model there are only the FRU units that can solve these issues.

So how can the same representations underlie effects with such different characteristics- they can not and this is clearly a limitation of the Bruce and Young (1986) model.

Question 56

what thus is needed in terms of models?

Answer 56

What is needed is a more tightly constrained model that can generate predictions about the patterns of reaction times.

Question 57

what other model was developed?

Answer 57

For this reason Burton Bruce and Johnston (1990) developed their Interactive Activation with competition model ‘IAC’- 1990,

Question 58

what are key differences between Bruce and Young model and Burton et al model?

Answer 58

• Bruce and Young unclear whether PINs held identity-specific semantic information or enabled access to this information. Burton et al., (1990) are clear that PINs provide an interface to semantic information (the SIUs), rather than holding semantic information themselves.
• This means that where a face, voice and other information are all provided, personal familiarity will be assessed from the combined activity arising from all these sources (rather than independently for face, voice, etc.).
• according to Burton et al., a face is seen as familiar when there is suffi cient activity at the PIN corresponding to that person’s face.

Question 59

how can the IAC model be used to explain semantic priming?

Answer 59

Activation increased at the Brad Pitt FRU when shown a picture of Brad Pitt. Clearly the Brad Pitt PIN crosses the threshold of recognition quite quickly, but Burton et al., 1990 showed that activation at the Angelina Jolie PIN also increases, but remains below the threshold. This happens because activation does not bring the Angeline PIN to threshold because it is simultaneously being inhibited by the more active Brad PIN.
After the input ceases activation at both PINs decays a little but does not return to resting level. This means that a subsequent presentation of Angelina or Brad will be recognised more quickly than if the activation in these PINS was at resting level.

Question 60

how does the IAC model explain the issue of multi domain information in semantic processing?

Answer 60

it will not matter whether this subsequent presentation comes from names or face because either input route will be able to capitalise on the residual increased activation level at the relevant PIN.
n Burton et al.’s (1990) IAC simulation, then, the mechanism for semantic priming lies in interaction of the PINs and SIUs, and the explanation of why semantic priming will cross input domains is because the PINs can receive input from FRUs or NIUs. The short duration of semantic priming is accounted for by the fact that subsequent presentation of any face (or name) other than Brad or Angelina will drive the Brad and Angelina PINs back to resting level, because of the within-pool inhibitory links.

Question 61

explain how the IAC model explains repetition priming

Answer 61

Whenever two units are simultaneously active the connection strength between them is increased. So, post recognising Brad Pitts face, the connection between the Brad FRU and the Angelina FRU will be strengthened, making activation of the PIN rise more quickly the next time Brad’s face is encountered.
Because of this strengthened connection, repetition of the face then leads to faster recognition. This explains why repetition priming effects are domain specific, however, because it is the links between input units (FRUs or NIUs) and PINs which are strengthened, so if the face has been seen the FRU-PIN link is strengthened but this buys no advantage when it comes to recognising the name later on, and vice versa.

Question 62

how can the IAC model also explain the long lasting effects of repetition priming

Answer 62

The time over which a strengthened link decays can be set to be short or long, as the modeller chooses, but if link strengthening is to be any realistic analogue of learning then slow decays are more plausible than rapid ones.