Visualizing Patterns
The Lovers and Art of Pablo Picasso
Objective:
To craft visuals that communicate data and highlight emerging patterns
Defining a Topic and an Interesting Question
Most of my initial ideas revolved around colours. One early idea was to look at dominant colours in paintings and map out colour trends across different art movements or by different painters.
The biggest challenge would be sorting the data — I’d have to try and find (+ teach myself how to use) scripts online that can scrape through metadata, or do it manually.
The other will be scoping the data, and doing a deep dive into a particular style, for example — only look at the renaissance in italy, or modernism ‚ or to do a comparison — impressionism vs post impressionism. This will be important especially if I end up having to sort through the data manually.
I thought an interesting topic would be to look at
How, if at all, have Picasso’s lovers affected his work?
Preliminary Data Collection
A paper by Roy C Saper has been my primary resource on Picasso’s major relationships. It appears to be the most definitive paper written on the subject, cited extensively by other writers. Other information I found confirms that the names of Picasso’s most notable lovers, as well as the duration of the relationships indicated in the paper are accurate.
I realised that while this list of lovers indicates no overlap, there are several accounts of Picasso having a mistress during all his relationships. I soon came to understand that the dates mentioned in the paper do not indicate the total period of Picasso’s engagement with each woman, rather the time during which his affections lay with her.
The major women in Picasso’s life have been:
1904–1912 Fernande Olivier
1912–1915 Eva Gouel
1917–1927 Olga Khokhlova
1927–1936 Marie-Therese Walter
1936–1944 Dora Maar
1943–1951 Francoise Gilot
1951–1953 Genevieve Laport
1953–1972 Jacqueline Roque
Finding a comprehensive resource for Picasso’s paintings is not an easy a task as it sounds. Picasso was famously prolific, having produced an estimated 1885 paintings over his 70ish years long career. At first I decided to look at this archive of selected paintings.
To look at colour trends across these paintings I used this script to identify a “dominant colour” in the image.
Defining questions
After pulling a preliminary sample of data, I discussed the questions I wanted to answer with some of my peers.
Is there a distinct colour palette from one woman to the next?
Are there more or less paintings dedicated to certain lovers?
Was he more or less prolific during certain relationships?
Is there any correlation between Picasso’s “Art Periods” and his lovers?
What subjects did he most paint while he was with a certain lover?
Collecting and Refining Data
After collecting images of approximately 200 paintings and identifying their dominant colours, it became apparent that though there were distinct colour palettes in different periods of Picasso’s life, they bore little correlation with the woman he was with at the time. Minor setback, if a setback at all.
I decided to investigate further into the relationship between Picasso’s lovers and his art. After watching several documentaries on the subject (and a surprising number exist—I recommend this one from the BBC) a really interesting pattern emerged: as the relationship progressed, Picasso’s depictions of the woman in his paintings changed visibly.
Many of Picasso’s works are portraits of the women he loved. I decided that it was more revealing to look specifically at the portraits he painted of his lovers, instead of paintings depicting all kinds of subjects. I found this gallery which has an acceptable collection of these portraits and provides separate pages for each of the women Picasso loved. I followed the same process of deriving the dominant colour of these paintings. I decided that the final piece should be able to not only convey colour trends to also allow the viewer to discover these changes in style.
Layering the Data
Before getting deep into the visualisation, I created a map to show how information would build up. This would guide not only interaction, but also visual hierarchy (starting from the baseline to more prominent information.)
This was a rough map with all the possible information I could show. I later ended up revising it, or only showing certain things since this was creating an information overload.
Visualizing the Data
Playing with scales and coordinate systems.
Initial brainstorming all revolved around time as the baseline for all other data categories, with time represented on a cartesian and radial coordinate system. Colour, texture and value did not work since colour is part of my data set, and using variables that interfere with the information confused things.
I needed to quickly move my visualisations into a digital medium where I could see the colours, and work with some realistic data in order to easily manipulate the form and see the resulting patterns.
In the above visualization, each block represents a painting, and its colour the dominant colour of that painting. The paintings in each year are arranged as they are listed in my data source, which I assume is chronological. However, there are no colour patterns that emerge clearly.
I rearranged all the paintings roughly according to colour instead of time (hue on the horizontal axis, saturation+brightness on the vertical axis). The result was interesting “colour profiles” that almost function as identities for each woman. It also gives a better comaprison of quantity as the area of each profile is much better visualised.
The problem with this visualisation does not indicate duration of the relationship in any way, since the two axes are already taken up by attributes of colour.
In order to show both time and colour, I needed to devise a colour scale which accounts for hue saturation and brightness, but on a single axis. After much trial and error, i was able to devise a relatively compact scale while still maintaining a wide enough range to fit the data.
Using colour on one axis and time on the other, I compiled the following viusalization:
While the scatter plot of colour seems visually interesting and reveals some patterns, I realised the biggest problem with this is that it goes against natural cognitive affinities. We tend to see time as moving left to right, not top to bottom. The horizontal bar graph doesn’t quite register the quantities.
Planning for Interaction
While preparing to present this piece to the class, I struggled with visualising all the data on one screen. Trying to show the big picture all together led to everything being too small to read, which was obviously a huge flaw. After getting the same feedback from the class, I realised that I had designed the visualisation as a static print piece rather than taking advantage of interactions to walk the user through the patterns and allow them to explore the data.
I created another map, with my parsed down set of data categories, to map out the interactions within the viz. This was very similar to the earlier map of layers, but only containing information that was relevant to my objectives.
I removed some of the data was not adding anything to the narrative I wanted to construct.
I also decided to switch back to showing time on a vertical top-to-bottom scale, realising that going through the years as you scroll down a page created a greater sense of the passage of time.
I added a highlighed section to the colour scale to allow users to explore painting belonging specific parts of the spectrum but not necessarily in the same time zone..
Final Viz
was presented as a keynote animation which can be accessed through the course drive folder.
