Correlation Lab Report – data to be run on SPSS using one tailed pearson, scattergraphs for each correlation to be included

Facial recognition; Does sensitivity to configure/featural differences in upright faces/houses predict self-reported face recognition ability?

Main Hypothesis we are testing:

  1. Configural sensitivity towards upright faces predicts self reported face recognition ability
  2. Featural sensitivity towards upright faces predicts self reported face recognition ability
  3. Configural sensitivity towards upright houses predicts self reported face recognition ability
  4. Featural sensitivity towards upright houses predicts self reported face recognition ability

The hypothesis is directional – One tailed as would expect a negative correlation as a high PI20 score corresponds with low self-reported face recognition ability. If not significant then why?

Data that needs to be analysed:

  • PI20 (questionnaire measuring self reported face recognition ability) scores (high scores indicate lower self reported face recognition ability and vice versa)
  • Experimental task -160 trials (data on sensitivity d)

-80 trials for each condition (ie 80 for upright faces, 80 for upright houses) of which:

-20 which differed in configuration but not features

-20 which differed in features but not configuration

-40 pairs which were identical

  • If there are significant correlations then partial correlation to be run to help explore further whether effects are upright face-specific, ie does sensitivity to configural differences in upright faces explain any of the variation of self-reported face recognition ability that sensitivity to configural differences in non face objects does not
    • Correlating d for configural differences in upright faces controlling for d for configural differences in upright houses
    • Correlating d for featural differences in upright faces controlling for d for featural differences in upright houses

Reason behind doing the experiement:

  • Previous research by Yovel & Kanwisher 2004  showed x, y, x but nor clear if a,b,c
  • Failed to look at 1,2,3 hence it was decided to reperform the experiment
  • Ie what hasn’t been answered by the previous resesarch
  • Want to see how reliable the previous research was

Background:

  • Evidence that people can recognize 5000 faces on average (but wide individual differences) – Jenkins et al., 2018
  • Recognition survives well if we haven’t seen someone for a long time, or they’ve changed their appearance (hairstyle etc)
  • People are also generally good at deriving other information from faces such as emotion, but again evidence of individual difference – Hoffman et al., 2010
  • How good we are is surprising as we see a lot of faces and their first order configuration (nose in the middle, eyes above, mouth below) is the same.
  • Individual features (nose, mouth, eye etc)
  • Second order configuration (spacing)
  • Holistic processing (integration of the multiple parts of a face into a single holistic representation)
  • See Maurer et al., 2002 for further discussion.
  • Many psychologists believe faces (particularly upright faces) and non-face objects are processed differently:
  • Range of effects (composite effect, inversion effect etc) found in behavioural experiments with faces are not found (or not to the same extent) with non-face objects – Robbins & McKone, 2007
  • Neuro-imaging studies show differences in activation (notably fusiform face area) – Kanwisher & Yovel, 2006
  • It has been proposed:
  • object processing involves decomposition into parts or features (Biederman, 1987)
  • faces are represented and recognised holistically (Tanaka & Farah, 2003) and in particular relying on second-order configuration (Searcy & Bartlett, 1996)
  • However, it has also been argued featural processing of faces has been underplayed:
  • emphasis on configural processing often relies on assumption that inversion primarily impairs configural processing, but evidence that it also impairs feature processing (Murphy & Cook, 2017)
  • Your experiment asks:
  • Does sensitivity to configure differences in upright faces predict self-reported face recognition ability?
  • Does sensitivity to feature differences in upright faces predict self-reported face recognition ability?
  • Is this pattern the same for upright houses?
  • Why is it relevant?
  • Face processing is important for social interactions & deficits could contribute to isolation etc
  • As well as prognosticator (face blindness) as an extreme form, various groups may have some difficulties with faces – autism (Dawson et al., 2005) and older people (Ortega & Phillips, 2007)
  • Could training help? If so, configure or feature, and would it be limited to faces or include other things?
  • Could configure and feature processing differences be a diagnostic tool?
  • Stimuli were houses and faces which differed either in features or configuration – Yovel & Kanwisher 2004
  • Stimuli were either upright or inverted, giving four conditions. This replicates Y&K but we will only give you, and you should only analyze, upright conditions
  • identical
  • sensitivity (d’) is a measure of accuracy which is independent of response bias
  • It takes into account both cases where you correctly saw there was a difference (“hits”) and where you correctly saw there was no difference (“correct rejections”)
  • Superior measure than just “hits” because it doesn’t matter in theory if you are biased towards or against reporting a difference.