Sutcliffe – October 2012 – p. 1 Research Proposal REQUIREMENTS

Sutcliffe – October 2012 – p. 1
Research Proposal REQUIREMENTS
Title:
The Iconology of Pinterest
Type of study:
This project will analyze the names created by user-curators when organizing social
image collections (Pinterest.com) The goal of this study is to determine to what
extent the created names can be assigned to Panofsky’s three tiers of meaning in the
visual arts. This project is not concerned with retrieval issues, but focuses on user
behavior when creating image collections, specifically when assigning meaning to
an image via the choice of names.
Problem statement:
Every image that is added to a user-curator’s collection in Pinterest requires two
user-curator-generated names: a board name and a pin name. How often do the
name selections of user-curators in Pinterest correspond to Panofsky’s tiers of preiconological (primary or natural), iconographical (secondary or conventional) and
iconographical (intrinsic or contextual) meaning? Applying Rosch’s levels of visual
categorization in addition to Shatford Layne's categories of image attributes will
provide a matrix necessary to quantitatively notate the levels of meaning available in
a given image name.
The purpose of this study is to analyze how often the image names selected by
Pinterest user-curators correspond to Panofsky’s three strata of subject matter
meaning (pre-iconographic, iconographic and iconological).
Purpose:
A cross section of random Pinterest entries will be collected. Each entry has two
user-curator assigned names: a board name (assigned entirely by the user although
Pinterest provides a set of default board names to adapt or ignore) and an individual
pin name (unique to that image.)
Each entry name (board and pin) will be assigned to one of Panofsky’s three strata
of subject matter and meaning.
To verify that the correct strata of meaning has been assigned, each pin name will
then be annotated using Rosch’s levels of visual categorization and Shatford
Layne’s nine categories of image attributes.
Significance:
User-curators may be developing particular sense-making behaviors as they actively
contribute names to large unstructured image collections. Understanding the
behaviors related to self-naming large numbers of images may lead to
improvements in user naming tools within other large social collection systems.
Do any specific factors encourage user-curators of a shared social collection to
contribute additional value to the collection in the form of meaningful image names?
Conversely, which factors may discourage user-curators from contributing to the
naming process?
Additionally, how well might a system developed in 1939 to interpret symbols in
Dutch oil paintings be expected to predict the naming selections of social image
collectors when creating large social image collections in the 21st century?
Sutcliffe – October 2012 – p. 2
Research Proposal REQUIREMENTS
Research question:
How frequently do the image names self-selected by user-curators in Pinterest
correspond to Panofsky’s three tiers of meaning: pre-iconological (primary or
natural), iconographical (secondary or conventional) and iconographical(intrinsic or
contextual) ?
Using Panofsky’s three levels as a lens, at what level do naming choices of usercurators of social image collections differ most widely from Panofsky’s system?
Where do the name choices correspond most closely to Panofsky’s three levels of
meaning?
Literature review:
Ways to think about image collections:
•
•
•
•
Greisdorf and O’Connor (2008) Structures of image collections.
Shatford Layne, S. (2002). Subject access to art images.
Hastings, S. K. (1999) Evaluation of image retrieval systems.
Shatford Layne, S. (1994) Some issues in the indexing of images.
Identifying meaning in images:
• Panofsky, Erwin. (1939) Studies in iconology.
• Beach, L. R. (1964) Cue probabilism and inference behavior.
• Panofsky, Erwin. (1972) Meaning in the visual arts.
• Reed, S. K. (1972) Pattern recognition and categorization.
• Tversky, S. (1977) Features of similarity.
• Rosch, Eleanor and Lloyd, B. (1978) Cognition and categorization .
• Mitchell, W.J.T. (1986) Iconology.
• Van Straten, R. (1986) Panofsky and ICONCLASS.
• Shatford, S. (1986). Analyzing the subject of a picture: a theoretical
approach.
• Mitchell, W J T. (1995) Interdisciplinarity and visual culture.
• Gombrich, E. H. (1999) The uses of images.
• Elkins, J. (1999) The domain of images.
• Shatford Layne, S. (2002) Introduction to art image access.
Naming images:
• Shatford, S. (1984) Describing a picture: A thousand words are seldom cost
effective.
• Hibler, D; Leung, C.H.C; Mwara, N. (1992) Picture retrieval by content
description.
• Jaimes, A. and Chang, S.-F. (2000) A conceptual framework for indexing
visual information at multiple levels.
• Schreiber, A.T; Dubbeldam, B; Wielemaker, J; Wielinga, B. (2001)
Ontology-based photo annotation.
• Hollink, L. et al. (2004) Classification of user image descriptions.
• Hanbury, A. (2008) A survey of methods for image annotation.
• Sandhaus, P. and Boll, S. (2010) Semantic analysis and retrieval in personal
and social photo collections.
• Rogiest, Peter. (2011) Introduction to controlled vocabularies: terminology
for art, architecture, and other cultural works.
Sutcliffe – October 2012 – p. 3
Research Proposal REQUIREMENTS
Data collection
method:
On a range of separate dates, a cross section of random Pinterest boards will be
downloaded. The names of the boards and the names of the included pins will be
compiled using content analysis.
Next, each name will be assigned to a matrix containing Rosch’s three levels of
image categorization and Shatford Layne’s four categories of image attributes.
Finally, each name will assigned to one of Panofsky’s three levels of subject matter
or meaning (pre-iconographic, iconographic or iconological.)
The study will collect and analyze XXX boards from XXX users over a time period
of XX days. In this project, power analysis suggests that the intended sample size of
176 boards is sufficient for a single-sample t-test yielding an effect size of 0.5
(medium as per Cohen’s d) with a power of 0.95
Pinterest users are anonymous in terms of reported demographic data so little
information related to age, gender, education or income can be deduced from normal
activity.
Pinterest was selected to exemplify social image collections in this project based on
the number of participants in 2011 and the increasing rate of growth in 2012.
Sutcliffe – October 2012 – p. 4
Research Proposal REQUIREMENTS
Data analysis method:
In this project, the relationship being measured is the correlation between the words
used in the names and Panofsky’s three strata of subject matter meaning (preiconographic, iconographic and iconological).
This project will use naturalistic observation, randomly recording the variables of
interest in the natural environment without interference or manipulation by the
experimenter.
A qualitative ethnological approach is planned, considering an entire group (the
universe of active pinners on the given sample dates) in a natural environment and
identifying the everyday information behaviors within the group (the naming of
random boards and pins created on those dates.)
Content analysis of a random selection of pin and board names will be used to
assign the collected names to one of three levels.
Pre-iconographical Description
Iconographical Description
Iconological Interpretation
Names which are factual,
recognizable and do not
indicate specialized
knowledge would have a
strong positive correlation
to Panofsky’s category of
Pre-iconographical
Description.
Names which rely on a
theme, a literary
allusion, specialized
knowledge, formulas,
allegories or other
layers of meaning
beyond the immediately
factual and recognizable
would have a strong
positive correlation to
Panofsky’s category of
Iconographical
Description.
Names which are
symbolic, culturally
specific, interpretive,
historically defined or are
non-contextually defined
have a strong positive
correlation to Panofsky’s
category of Iconological
Interpretation.
There are three possible results of a correlational study: a positive correlation, a
negative correlation, and no correlation. The correlation coefficient is a measure of
correlation strength and can range from –1.00 to +1.00.
Positive Correlations: Both variables increase or decrease at the same time. A
correlation coefficient close to +1.00 indicates a strong positive correlation
Negative Correlations: Indicates that as the amount of one variable increases, the
other decreases (and vice versa). A correlation coefficient close to -1.00 indicates a
strong negative correlation.
No Correlation: Indicates no relationship between the two variables. A correlation
coefficient of 0 indicates no correlation.
Sutcliffe – October 2012 – p. 5
Research Proposal REQUIREMENTS
The practical significance of this study will be assessed qualitatively, by compiling
the names from random selections and assigning them using content analysis to
Panofsky’s three tiers. The resulting research report will include the specific names
tabulated, a thesaurus of terminology used during content analysis, the rating
procedure used to determine suitability of names to their tiers, reported frequencies
for each name and a description of found data patterns.
Methodological issues: Correaltional studies can be time consuming.
Naturalistic observation does not allow for scientific control of variables.
Control for extraneous variables may not be possible.
How to determine statistical significance ?
[A single sample t-test with a ratio less than .05 is considered significant. ]
Scope and limitations:
Selecting effectively random samples of pins without retrieving unmanageable
numbers of names may require several pilots.
Pinterest users are anonymous in terms of reported demographic data so little
information related to age, gender, education or income can be deduced from
categorization activity.
Terms and
definitions:
While correlational studies can suggest that there is a relationship between two
variables, they cannot prove that one variable causes a change in another variable.
The research design of this study does not completely control for bias and the effects
of other variables. This may limit conclusions regarding causation.
Rosch’s three levels of image categorization:
(1) Subordinate images share very few attributes (1996 Volkswagen Beetle)
(2) Basic images share attributes common to most members of the category (car)
(3) Superordinate images share only a few attributes (transportation)
“Very few attributes are usually listed for superordinate categories. Significantly
greater numbers of attributes are assigned to basic level objects. Subordinate level
objects do not have significantly more attributes assigned than do basic-level
objects.”
Shatford Layne’s four categories of image attributes:
 can cover of- ness ( a picture of a dog)
 can cover about- ness (a picture about loyalty)
 can be generic (animals)
 can be specific (2-year-old Great Dane female with a red collar)
Panofsky’s three strata of subject matter or meaning include:
 pre-iconological (primary or natural)
 iconographical (secondary or conventional)
 iconographical (intrinsic or contextual)
Sutcliffe – October 2012 – p. 6
Research Proposal REQUIREMENTS
Expected results:
Based on observation, the expected results should show that both pin names and
board names selected by user-curators in Pinterest correspond to all three levels of
Panofsky’s strata of subject matter and meaning.
It is expected that names which are factual, recognizable and do not indicate
specialized knowledge (the strongest positive correlation to Panofsky’s category of
Pre-iconographical Description) would occur most often in pin names.
Completion schedule:
It is expected that names which rely on a theme, a literary allusion, specialized
knowledge, formulas, allegories or other layers of meaning beyond the immediately
factual and recognizable or include symbolic, cuturally specific, interpretive,
historically defined or non-contextually defined words (the strongest positive
correlation to Panofsky’s category of Iconographical Description or Iconological
Interpretation) would occur most often in board names.
Pilot data collection – Fall 2012
Collect and analyze data – Spring 2013
Write report – Fall 2014