Face Recognition in Applied Contexts - Lecture 4

Question 1

Who thought identification contributed to face recognition?

Answer 1

Burton, Bruce and Hancock, 1999

Question 2

Who thought gender contributed to face recognition?

Answer 2

Bruce and Langton, 1994

Question 3

Who thought attractiveness contributed to face recognition?

Answer 3

Fleishman et al 1976

Question 4

Who thought emotional expression contributed to face recognition?

Answer 4

Calder, Burton, Miller, Young and Akamatsu, 2001

Question 5

What are the four things separate from object recognition?

Answer 5

1. Faces are dynamic
2. Large research area (Young, 1998)
3. Different neurological pathways (Blonder et al, 2004)
4. Prosopagnosia (Morrison, Bruce and Burton, 2001)

Question 6

Identification of familiar faces is….

Answer 6

generally very good under various conditions (Hancock, Bruce and Burton, 2000).

Question 7

What are the four conditions?

Answer 7

Lighting changes
Disguises
Viewpoints
Expressions

Question 8

Problems in unfamiliar face recognition. What are the imaging problems?

Answer 8

1. Research has focused on frontal views of faces but faces are 3D and complex.
2. 3/4 view is best recognised - partly because this lies between frontal and profile so any change is relatively small (O’Toole et al, 1998)
3. Profile is particularly bad especially when generalising from one profile to another (Hill et al, 1997).
4. Profiles obscure much of the configurable information that seems to be important.
5. Inverting a face also makes recognition difficult.

Question 9

What did Bruce and Langton (1994) say about negatives?

Answer 9

That they are very hard to recognise but they still provide lots of information like position and size of facial features.

Question 10

What must representations do?

Answer 10

They must encode more information than negatives from the original image, hence the difficulty.

Question 11

What is it that causes difficulty in recognising negatives?

Answer 11

The loss of shading (Hayes et al, 1986)

Question 12

What happens if you light an image from below?

Answer 12

It has a similar effect to negation and it disrupts identification (Johnson et al, 1992).

Question 13

What did Hill and Bruce, 1996 reveal about bottom lighting a photo?

Answer 13

That even lighting one photo from the bottom, even if the 2 photos are from the same viewpoint, makes matching hard.

Question 14

What does inverting a face do?

Answer 14

It makes viewers less sensitive to configural information compared to upright faces.
It causes a loss of configural information
‘Thatcher illusion (Thompson, 1980)
Inversion effect = greatest with faces compared to houses or other objects.

Question 15

What did Bartlett and Searcy, (1993) do?

Answer 15

They made unreasonable configural adjustments. However, ‘grotesqueness’ ratings were much lower when these faces were inverted.

Question 16

How does distinctiveness effect recognition?

Answer 16

1. Distinctive faces are most likely to be remembered (Light et al, 1979).
2. Faces are recognised more quickly if familiar and distinctive (Valentine and Bruce, 1986)
3. Caricatures are recognised more quickly than ‘true drawings’ (Rhodes et al, 1987)
4. Caricatured drawings may be rated as a better likeness of the face than the actual photograph (Benson and Perrett, 1991)

Question 17

What 6 things influence recognition?

Answer 17

1. Distinctiveness/familiarity
2. Disguises - covering the hair hair (Cutler et al, 1987) and sunglasses (Hockley et al 1999) because they’re important in recognition.
3. Wells and Olson (2003) note that light levels should influence recognition and hence eyewitness testimony.
4. Length of exposure to culprit (Ellis et al, 1977) although ‘weapon focus effect’ (Loftus et al, 1987).
5. If abstract judgements are made, then recall of face is better (Wells and Hryciw, 1984).
6. Changes over time - facial hair/ageing.

Question 18

Does gender alter effectiveness of witness?

Answer 18

Small effect and overall a very little difference (Shapiro and Penrod, 1986)
Some advantages for females (Megreya et al, 2011)

Question 19

Does age alter effectiveness of witness?

Answer 19

Young and old make errors - mainly when culprit is absent (Pozzulo and Lindsay, 1998)

Question 20

Does level of intelligence alter effectiveness of witness?

Answer 20

Howells (1938) found results suggesting so but since then very little effect (Brown, 1977).

Question 21

Does race alter effectiveness of witness?

Answer 21

Better identification of own race (Meissner and Brigham, 2001) known as ‘other effect’ or ‘other race effect’.

Question 22

Does personality alter effectiveness of witness?

Answer 22

Little evidence to suggest it does although highly anxious make less mistakes (Shapiro and Penrod, 1986) - could be that they make fewer identifications overall.

Question 23

What are the 4 problems with traditional measures of retrieval?

Answer 23

1. Compositions/composites vary in likeness (Ellis et al, 1975)
2. Mug shots can cause interference (Deffenbacher et al, 2006)
3. Identification parades can be biased (Buckhour et al, 1975)
4. CCTV is a solution for these problems. However, such images are far from ideal. They may be very small, poor quality, take from strange angles or badly lit - makes matching hard. Even good quality video images can be hard to identify with unfamiliar faces (Burton et al, 1999).

Question 24

What did Henderson, Bruce and Burton (2001) reveal?

Answer 24

Participants were required to match CCTV images to a face from a choice of 8 mug shots.
First and second choice mean accuracy was 28.5%
Some faces chosen more than actual robbers.
Matching photo to photo improves the mean to 76% for robber 1 and 33% for robber 1.
Further experiments used broadcast quality pictures and varied whether robbers were disguised or not (wearing a hat).
Presence of disguise reduced accuracy to 42.5% from 63.5%
Reduced to just 2 photos and no hat and error rate was still 35% in some cases.
Overall, CCTv may be helpful in many ways, recognition of unfamiliar faces is hard.

Question 25

Can training be effective?

Answer 25

Wilkinson and Evans (2009) compared face imagery experts to the public on CCTV matching images to photos - double accuracy rate and half error rate.
Megreya, Sandford and Burton (2013) found increased error rates when images from different days were matched, suggesting underestimates of difficulty.

Question 26

Does length of exposure help with identification?

Answer 26

Memon and Bartlett (2002) reported that older adults (60-80) more prone to identification errors.
Memon et al (2003) used 2 (exposure lengths) x 2 (line up type) x 2 (age group) design.
Longer exposure did yield higher accuracy and even higher when target present than absent.
No effect of age.
Weapon effect and type of judgements made.

Question 27

Is memory for actions better than for faces?

Answer 27

Yuille and Cutshall (1986) assessed accuracy by attempting to reconstruct an event via various sources of information - found higher accuracy and reliability.
Woolnough and Macleod (2001) compared statements with actual CCTV footage of crimes - true ‘objective assessment’ of eye witness accuracy.
Good recall for violent crimes - attracts and directs attention - perhaps just to activity rather than the perpetrator?
Both victims and bystanders had accurate memory for events.
96% accuracy rates - are they just flashbulb memories?
Tollestrup et al (1994) - fraud victims act like lab participants with rapid forgetting, little info and poor identification.