Research into Visual Perception conducted by Barnard
Following the failure of the early investigators into visual perception
of simple two-dimensional shapes to reach conclusions consistent with
each other, and to explain their findings systematically,
experiments of the same form were still being undertaken twenty years
and Fehrer’s account of the rôle of form in perception.
See for example, work by Blackwell, Casperson, and Aulhorn.
In 1972 Barnard's work exploring the possibility of a new image
evaluation method based on target recognition statistics was reported.
This investigation was
at a time when the images generated were produced on photographic
film, rather than on CRT T.V. monitors (as was usual in the 1980s and
1990s) or LCD displays (which has a technical and affordable option
since about 2000).
Barnard summarised the purpose of his
“ The selection of an image evaluation method
has several important consequences. It establishes a basis for measuring
the performance of
imaging systems, and it implicitly defines criteria for the optimal
design of such systems. Since the image evaluation method influences
design, it must be properly matched to the intended system utilization
involves the conveyance of visual information to human observers.”
That is, this work was conducted as part of "human factors" research
develop improved engineering systems - rather than purely to better understand
Barnard’s study was based on experiments in which arrays of simple
shapes (including, for example, the Landolt Ring and a figure
with five ‘legs’ 60 degrees apart) were
presented to observers. In each case the target symbols were positioned at
random orientations and arranged so that the size of the symbols decreased
down the page.
Examples are shown below:
The method used for these experiments differed from the earlier (and
many subsequent) works because in this case the observers arranged their
viewing conditions, such as viewing distance and duration, with complete
freedom. It was noted that the observers participating in this experiment
consistently performed better than they expected. This observation was
taken to indicate that ‘people extract and use information of which
they are not consciously aware ’.
Barnard analysed his results by plotting curves of probability (of correct
specification of Landolt Ring orientation; orientation of missing spoke;
recognition of numerals) against the reciprocal of target size. This assumption
that probabilities, rather than absolute thresholds, should be analysed
when studying human perception is also noteworthy.
For further information about this work see Barnard's Paper:
T.W.Barnard, "Image Evaluation by Means of Target Recognition",
Phot.Sci.and Eng. Vol.16, pp.144-150 (1972).