In Progress

Perhaps good ideas replicate themselves. Just a few days after this site suggested that the campaign to hack the Eighth Amendment was given a lift by Justice Kennedy’s recent testimony before the House, the New York Times editorial board published a great piece invoking Kennedy’s testimony and calling for the Court to end long-term isolation because it violates that same Amendment. Even better, the Times edit specifically praised  Kennedy’s movement away from a weak Eighth Amendment view in the 2003 California Three-Strikes cases to his more robust perspective on what constitutes “cruel and unusual punishment” in more recent cases involving the death penalty, juvenile life without parole, and prison overcrowding.

Want a better way to access what the NYT was talking about? Try this:

The map above links via Casetext to all the opinions mentioned by the NYT in its 4/4/15 editorial. Please note that this is a “standard view” map — the Y-axis gives the number of votes an opinion receives. Thus the first two opinions on the map, from Breyer in Ewing and Souter in Lockyer are dissents — they received 4 votes. (Technical note: Ewing and Lockyer were actually both decided on 3/5/2003; I changed Lockyer‘s date to 3/6/03 so that it can be better viewed the map).

In online version of their editorial, the NYT provided links to the Oyez and LII versions of the particular cases. Of course, both of these free services are fantastic and deserve our support. One particular advantage to linking to the Casetext versions, however, is that it allows users to annotate and provide commentary on cases in a wiki-like fashion. Thus, if readers want to highlight key language from these NYT-approved cases, they can do so. Hack away!

The NYT edit call

Last week, I taught the Flag Desecration Cases in my First Amendment class. It’s an interesting closed citation network since it contains only five cases. (Of course, the First Amendment doctrine implicated by these cases extends widely, but there are only five cases where the Court has directly considered criminal convictions for flag desecration.) This kind of tiny network provides an easy case study to compare the two main types of visualization employed here at the SCOTUS Mapping Project.

 The map above shows all five cases using a Spaeth projection. As I’ve explained before, this projection displays data from the Supreme Court Database (commonly known as Spaeth). Vote for outcome (5-4, 6-3, etc) are displayed on the Y axis as well as outcome type (“liberal” or “conservative” as coded by Spaeth).

This network is remarkable in that EVERY case shown is coded liberal by Spaeth. This is because in every flag desecration case that has come before the Court, the conviction has been vacated on First Amendment grounds. At the same time, this result has been consistently controversial. The overall “degree of dissent” in the network is a whopping 0.9 — close to an average 5-4 vote. (Recall a 9-0 case has a 0.0 degree of dissent; a 5-4 has a 1.0 degree of dissent).

Now — compare the map above with this one below, which uses a different projection.

This map shows the exact same territory but uses a standard projection (so-called, because this was the original projection used by the SCOTUS Mapping Project). In this view, the Y axis represents the number of votes an OPINION receives. Thus, every opinion above the dashed middle line is a majority opinion. This projection also gives the name of the author below the case name.

Note how the standard projection presents a slightly different view of what was going on in Goguen and Spence. Both of those cases had 6-3 votes for outcome (as shown in the Spaeth map), but the standard map shows that majority opinions in both cases garnered only five votes. Why? Justice White concurred in judgment only in Goguen while Justice Blackmun concurred in judgment only in Spence. The standard projection better  depicts such doctrinal subtleties.

The standard projection also showcases dissents and provides the names of authors of opinions. In this sense, it is richer. However, standard projections cannot — as of now — be generated automatically. Standard maps must be put together by hand using the software. Spaeth projections, on the other hand, can be automatically created by the app. As a means of taking a “first pass” at spotting the competing lines in an area of doctrine, I  find Spaeth to be incredibly useful.

Though the standard projections cannot be automatically generated, I have found that integrating Casetext into the process speeds things up. If you click on either map above, you’ll notice that the opinions are in turn linked back to Casetext. As I set about creating the standard projection, I used the Casetext “Quick Facts” features repeatedly to get my data about the individual opinions constituting the map. I then annotated the Casetext cases and linked those annotations back to my map. The result is a fairly robust and self-referential set of dissent-sensitive links.

This mode of linking to user-determined points within an opinion is a first here at “In Progress” and one that strikes me as potentially very useful for folks working with closed networks of Supreme Court doctrine. If any reader is interested in creating such a Casetext Map for their own research or teaching, please let me know.

Two weeks ago, Slate published an excellent interview with Berkeley law professor Jonathan Simon, author of Mass Incarceration on Trial. In the interview, Simon characterizes the fight against mass incarceration as a moral imperative and suggests that lawsuits have a role to play in the fight. What is needed, Simon suggests, is “a new generation of federal judges” who are willing to act on “the moral sensibility” that Justice Kennedy’s majority decision in Brown v. Plata has “reinvigorated.” Amen!

For those not familiar with Plata, it is the 2011 case where the majority (per Kennedy) held that severe overcrowding in California’s prisons constituted cruel and unusual punishment. Kennedy’s opinion has a moral authority and it is clear that the case affected him personally. Thus, in testimony last week before the House, Justice Kennedy stated that “the idea of total incarceration just isn’t working” and argued that solitary confinement “literally drives men mad.”

In his Slate interview, Simon compellingly suggests that Plata creates a “potentially safe doctrinal space” for future Eighth Amendment lawsuits. This seems absolutely right. Yet I hope that Plata is not the only case occupying this doctrinal space. Brave federal judges will need more than a single opinion to hang their hat on. How can we discover other opinions that might inspire moral action? Perhaps crowdsourcing Eighth Amendment doctrine can help.

Last month, I threw out the idea of hacking mass incarceration based on Plata. However, I’ll admit that I failed to offer a concrete way for others to plug into the effort. Now I have a new idea. It involves leveraging an awesome free legal resource — Casetext. The Casetext platform allows users to mark up and annotate cases in wiki-like fashion. My idea is to create maps of Eighth Amendment cases linked to Casetext and encourage folks to seek out and mark up useful Eighth Amendment arguments within those cases. As an initial experiment, take a look at this map. If you click on the image above, a full-sized version of the map will open in a separate window. Then, if you click on any given case, you’ll be taken to the Casetext version of the opinion. From there, you can mark up and annotate the case in any way you see fit.

Now careful readers will notice that the map above does not include Plata. Rather, the map represents the 2-degree network linking 2003’s Ewing v. California back to 1980’s Rummel v. Estelle. I wanted to start this experiment with this particular network for three reasons. First, it is a small and manageable network — only 7 cases. Folks can get up to speed quickly if they have interest. Second, the doctrinal line directly implicates mass incarceration. Rummel and Ewing both upheld obscene and excessive sentences against Eighth Amendment challenges. This map thus represents doctrine that must be undermined — at least those cases in red must be undermined — if sentencing practice is to be successfully confronted in the courts.

Third and finally, this is a deeply divided network. Five of the seven cases are 5-4s while the other two are 6-3s. (The degree of dissent [DOD] is therefore 0.93 — a network of all 9-0 cases would have a 0.0 DOD whereas a network of all 5-4 cases would have a DOD of 1). In my view, this high degree of dissent means that the doctrine is potentially unstable and susceptible to change. Just as importantly, there could be some potentially useful dissents that suggest an alternative vision for the jurisprudence.

In future posts, I will link this network up to Plata as well as back to earlier cases. I will also start mapping key dissents and concurrences separately. At this point, however, I invite folks to look at these cases on Casetext and share their thoughts and insights there. If this experiment works, we can tweak the project in whatever way works according to the wisdom the crowd.

Earlier this week, I looked at the challenges of editing a Supreme Court citation network through the lens of the First Amendment line of cases pertaining to commercial speech. Today I want to fill in the doctrinal picture just a little and highlight a different kind of doctrinal map. Specifically, I use a standard vote-based map to illustrate how the doctrine-altering Virginia Pharmacy Board (1976) majority decision emerged from prior dissents.

By way of background, know that purely commercial speech (i.e. advertising) lacked explicit First Amendment protection until the Court’s Virginia Pharmacy Board decision.  Prior to that, the non-protected status of commercial speech has been affirmed in a line of cases stretching back to a unanimous 1943 case called Valentine v. Chrestensen. So how did the doctrine change? Here’s a visual explanation:

The map above uses a “standard” display where the Y axis shows the number of votes a particular opinion receives. Unlike the Spaeth visualizations in my last post, this standard display makes room for dissents and concurrences and names the authors of opinions. As a result, the standard display better reveals the dialectics in Supreme Court doctrine.

The map illustrates that the Court’s formal decision to grant commercial speech First Amendment protection in Virginia Pharmacy Board was presaged by a series of dissents and concurrences. Justice Douglas, a member of the original unanimous majority in Valentine expressed doubt about that case’s viability in a 1959 concurrence (Cammarano).  The turning-point case was Pittsburgh Press Co. v. Pittsburgh Commission on Human Relations (1973). The unprotected nature of commercial speech led to a result in that case the dissenters regarded as deeply problematic. The loud calls for reform by the dissenters eventually persuaded the rest of the Court except for then-Justice Rehnquist. As the maps shows, this dialectic continues. Justice Rehnquist kept flying a flag against strong protection for commercial speech in Central Hudson (1980) and this flag was lifted again by the dissenters in 2011’s Sorrell decision.

Based on the reactions of my students, this standard visualization is easier to grasp than the Spaeth visualization. This makes sense to me — the Y-axis is certainly easier to interpret and the overall arc of the doctrine is easier to follow. While I too generally prefer standard visualizations, it is important to realize that, at this stage of development, they are far more labor-intensive to produce. By their very nature, Spaeth visualizations leverage the coding done by the Supreme Court Database team. Citations are also automatically generated using the CourtListener API. By contrast, creating standard maps requires the user (usually me!) to read the cases individually and tease out the relevant cites. I enter the data into the software manually and it then creates the visualization.

Of course, it is also possible to generate some of the map automatically and then manually enter additional information. This is exactly what I did with all the cases from Virginia Pharmacy Board onward. Once again, the lesson here is that while computer-assisted network analysis is extremely valuable and efficient, it cannot yet entirely substitute for human reading and editing. The level of complexity in Supreme Court doctrine is simply beyond the capacity of current algorithms to completely capture.

This semester I’m teaching First Amendment law for the very first time and it’s proving to be a wonderful experience. First Amendment cases have great stories behind them and the Supreme Court’s doctrine is complex and highly contested. To navigate this complexity, I’ve found doctrinal mapping enormously useful for my own learning. Yet I’ve also realized that automatically generated citation networks need editing. Today I want to discuss this editing process.

The example I’ll use is the Court’s “commercial speech” doctrine, which generally concerns acceptable versus unacceptable restrictions on advertising. The section on commercial speech spans 27 pages of our class textbook — the excellent (imho) Sullivan & Feldman First Amendment volume (5th Ed. 2013) — and it discusses 23 cases. There are three principal cases (Virginia Pharmacy (1976), Central Hudson (1980), and 44 Liquormart (1996)) and 20 squibs. The latest squib is the only Roberts Court decision — Sorrell (2011). For the purposes of this analysis, I’ll call these 23 cases the Sullivan & Feldman canonical cases.

To create a machine-made competing map, I used SCOTUS Mapper to generate a 2-degree network linking Sorrell to Virginia Pharmacy. The program’s algorithm pulls into the network all the cases that Sorrell cites that in turn cite Virginia Pharmacy. This is what that network looks like (as with all the images in this post, click for full-size map).

 Now this 2-degree network is actually quite rich. It contains 25 cases — including all three of the Sullivan & Feldman principal cases as well as 11 of the 20 squibs. While the 2-degree network thus picks up 14 out of the 23 canonical cases, it also picks up 11 “extra” cases not included in the canonical line.

I wanted to edit out these extra cases so I ran the network through a text filter on “commercial speech.” This machine-based edit knocked out 4 cases (the remaining 21 cases all contained the phrase). From there, I had to edit out by hand the 7 cases that Sullivan & Feldman did not include in the canonical line. (Let’s call those cases “non-canonical 7” — I’ll return to them below). After that editing, I ended up with this map:

Note that this second map uses a “Spaeth projection.” The Y axis is no longer random — it represents that Supreme Court Database code for both outcome direction and judgment vote. Red cases are Spaeth-coded “conservative” — meaning the Court upheld a restriction on commercial speech. Blue cases are Spaeth-coded “liberal” — meaning the Court struck down a restriction on commercial speech.

Now what does it take to automatically capture the remaining 9 squib cases from the Sullivan & Feldman canonical line? I tried generating a 3-degree network connecting Sorell to Virginia Pharmacy. So in addition to all the 2-degree cases, this network includes all the cases cited by 2-degree cases that in turn cite Virginia Pharmacy. Here’s what that network looks like on a random Y-axis projection:

Unsurprisingly, this network is very large — 81 cases. The good news is that the network easily picks up all 23 cases from the Sullivan & Feldman canonical line. The bad news is that it picks up 58 extra cases.

To get rid of these, I first tried the machine filter route. After automatically excluding cases without the phrase “commercial speech”, the network shrunk from 81 to 53 cases. Once again, that’s a good start but not nearly good enough. To get rid of the other 20 cases, I had to edit by hand. Upon completion of that edit, here’s what the new network looks like using a Spaeth projection:

This map represents all the Sullivan & Feldman canonical commercial speech cases. Getting those extra 9 cases in requires a healthy dose of editing of the 3-degree network.

So how important are those extra 9 cases? Are they really part of the core commercial speech line? All that we can conclude for certain is that none of the opinions in Sorrell cited those 9 cases. As far as the justices sitting in 2011 were concerned, none of those 9 cases was important in justifying their decisions. Of course, the diminished “2011 value” of the cases does not mean that “the missing 9” had no impact on the network’s development.

Comparing the 2- and 3-degree edited maps, one feature jumped out at me: almost half of the canonical cases missed by the 2-degree map concerned lawyer advertising (4 out of the 9 missed cases).  With a little more editing, I modified the last map to highlight the lawyer advertising cases in magenta. This is the result:

With the benefit of this new visualization, we can easily appreciate how the Sullivan & Feldman text took a deeper academic dive into the lawyer advertising cases that the Sorrell opinions found necessary. It seems hard to fault Sorell for only citing 3 of the 7 cases in the magenta line.

The other 5 missed cases — Carey (1977), Metromedia (1981), Posadas (1986), United Reporting (1999), and Lorillard (2001) — are certainly important. But are they more important than the “non-canonical 7” cases referred to above? Recall those are the 7 cases included in the 2-degree network after applying the “commercial speech” filter. In other words, those are ostensibly commercial speech cases cited in Sorrell but not included in Sullivan & Feldman.

Forgive me as I dive deep into the First Amendment weeds for just a moment more and name the “non-canonical 7” cases: Dun & Bradstreet (1985), RAV (1992), Edenfield (1993), Glickman (1997), Greater New Orleans Broadcasting (1999), Playboy (2000), and United Foods (2001). Three of those cases (Dun & Bradstreet, RAV, and Playboy) certainly do not belong the main commercial speech line. But the other four do. And they are arguably as important as the 5 missed cases above.

Now let’s step back and review. Sullivan & Feldman have 23 cases in their (human created) canonical network.  The machine generated 2-degree network (filtered for commercial speech) has 21 cases. 14 cases overlap between the networks and are clearly core to the line. Of the 9 cases captured by Sullivan & Feldman but not the machine, all are relevant and 5 are uniquely so. Of the 7 cases captured by the machine but not by Sullivan & Feldman, 3 are irrelevant and 4 are uniquely relevant.

All in all — Sullivan and Feldman’s editing fares better. This is as you would hope and expect. But the 2-degree network is still remarkably efficient at identifying relevant cases. (The 3-degree network, on the other hand, is far too large and unwieldy.) And the machine-generate network approach suggests potentially fruitful doctrinal angles for further reading outside of the Sullivan & Feldman line.

In the end, it bears emphasis that identifying relevant cases is very different from reading and understanding those cases. And in that department, the human editing of a casebook is completely indispensable. It would probably take 100s of extra hours to read unedited versions of cases identified by the 2-degree network. So thank your stars for human editors!

Hacking complex systems can improve them. White-hat hackers can open up needlessly closed systems, knock down senseless limitations, and then make connections and create solutions that improve that system for everyone. So here’s a thought: our system of mass incarceration needs some white-hat hacking. It serves no one well and breeds violence and misery in our society.

Luckily, conversations are happening. Just last week, the MacArthur foundation announced a $75 million “long-term strategy of investment in local reform, research, experimentation, and communications aimed at addressing over-incarceration by changing the way America thinks about and uses jails.” Conservatives and liberals now admit we have a problem and so the climate is right for hacking this problem.

Here at In Progress, we hope to contribute to collective efforts by hacking Supreme Court doctrine. The idea is to open up the law around prisons and make connections to help generate anti-mass incarceration constitutional arguments. What’s more, the goal is to crowdsource the connection-making process. To do this, I am experimenting with doctrinal map designs to facilitate non-specialist learning of complex doctrinal systems.

Last time, I charted out a series of Eighth Amendment doctrinal networks and found them large and unwieldy. It’s not realistic to expect anybody to read over 100 Supreme Court cases while mining for anti-mass-incarceration arguments. So this time, I want to narrow the focus. Below find the 2-dgree citation network linking the Court’s 2011 prisoner overcrowding decision Brown v. Plata to 1958’s Trop v. Dulles, a seminal pronouncement about the Eighth Amendment’s meaning as a guarantee of human dignity in light of evolving standards of decency.

Note the new design feature of the map above: its interactivity. Click on the map and then click on any of the opinions. You’ll find yourself looking an HTML deck that (a) has a very quick summary of the case holding; and (b) contains links to open resources about the case provided by CourtListener, Cornell Legal Information Institute, Oyez, and the Supreme Court Database. Some of the decks also contain other potentially useful information — check out the Brown v. Plata deck as an example (make sure you tap your right arrow key!).

As the above map demonstrates, legal hacking is a collective activity. If the map helps at all, it is only because it leverages free resources provided by great organizations doing great work. The HTML Deck platform is an especially cool free resource created by Dave Zvenyach, the 2014 DC Legal Hacker of the Year. His example should inspire us all to tinker and build and seek creative solutions.

Before signing off this post, let me return to the mass incarceration problem. While the map above does not provide any answers or pre-packaged strategies for ending this national problem, it does suggest an initial research agenda for putative doctrinal hackers. Technology won’t do the reading for you, but it can making your reading about the Supreme Court and prisons more directed and efficient. Does Eighth Amendment doctrine contain good arguments against mass incarceration? Maybe this map can help you look.

Last week, the Baltimore Sun published an editorial I wrote linking the phenomenon of exonerations to the problem of mass incarceration. Our justice system is overwhelmed by sheer volume, I argued, making wrongful convictions inevitable. I concluded by suggesting that we consider a radical solution to the mass incarceration problem — concerted policy of decarceration.

This suggestion obviously raises many difficult questions. Most such questions — What would decarceration look like? What is the right balance between freedom and safety? — have nothing to do with this blog’s domain of doctrinal mapping. However, over my next few posts, I want to examine a question that does resonate here: Does Supreme Court doctrine contain any potential constitutional arguments against mass incarceration? Is there a way mapping could help us find such arguments?

It’s undoubtedly a long-shot. Mass incarceration is a highly complex phenomenon crossing multiple jurisdictional lines and so it won’t likely reduce to a single case or controversy. Yet it also strikes me as unduly cynical to believe that the constitution has nothing to say about one of the most troubling social realities of our day. No matter how you interpret our constitutional commitment to liberty, locking up over 2 million souls in prison should seem suspect. Maybe then, just maybe, there are kernels of viable arguments against mass incarceration in Supreme Court doctrine.

How to start? Brown v. Plata is as good a place as any. In this 2011 case, the Supreme Court held 5-4 that a court-mandated population limit was necessary to remedy a violation of prisoners’ Eighth Amendment rights. California prisons were (and are) severely overcrowded and the Court upheld the decarceration policy — mandatory release of 38,000-46,000 inmates — fashioned by the lower federal court as consistent with the Prison Litigation Reform Act (PLRA).

Although the PLRA was at issue, the Eighth Amendment is where the action against mass incarceration lies. The idea is that overcrowding is cruel and unusual. How far back does that idea go. In his majority opinion, Justice Kennedy opened his Eighth Amendment analysis stating: “The basic concept underlying the Eighth Amendment is nothing less than the dignity of man.” For this hopeful proposition, Kennedy cited back to Trop v. Dulles (1958). This makes a good endpoint for a network.

So let’s begin this project by constructing a citation network linking Brown v. Plata to Trop v. Dulles. Using the SCOTUS Mapper software, I identified the 2 degree connections (cases that Plata cites that in turn cite Trop), 3 degree connections (cases that cited/are cited by by 2-degree cases that in turn cite Trop) and 4 degree connections. There were 9 2-degree cases, 48 3-degree cases, and 186 4-degree cases. (Click on the links to see a schematic visualization of those networks).

Now 186 4-degree cases is WAY to many cases to look at — and 48 3-degree cases is also a bit too demanding. So using CourtListener’s search tools, I filtered the networks above to only display those cases that contained the words “prison” (and related words like “prisoner”) and “Eighth” (here I did an exact search). This had the effect of seriously reducing the size of the networks – though the 4- and 3-degree are still large.

186 4-degree cases was filtered down to 77 cases — a 59% decrease. (Click on image to for a larger view).

48 3-degree cases was filtered down to 33 cases — a  31% decrease. (Click on image to for a larger view).

Finally, the 9 2-degree cases were filtered down to 8 cases. This is a mere 11% decrease in network size and it suggests, as is intuitive, that the 2-degree cases are most directly related to each other in content around prison and the Eighth Amendment. Interestingly though, this filter knocked out Trop v. Dulles itself. Although it provides staple quotes about the Eighth amendment in the prison context, that case was not about prisons.

In my next post, I’ll begin the process of examining the cases in the networks above and trying to observe patterns. For now though, I’ll leave readers to explore the networks above on their own and see if anything useful jumps out!

Over the weekend, law student and PhD candidate Ryan Whalen posted a fascinating map of the Law Prof Twitter Network. Ryan created the map by crunching data compiled by Prof. Bridget Crawford over at the Faculty Lounge. The image is massive, striking and thoroughly fun to explore. In fact, the map provided such infectious fun that it inspired me to experiment with creating a Very Large Map of my own. Today I present the results of that experiment. It constitutes the third look in a series of three looks at Supreme Court’s fighting words/hostile audiences doctrine.

As I’ve previously discussed, I am teaching First Amendment law this semester using the  Sullivan & Feldman textbook. They devote Section 3 of Chapter 1 to “Fighting Words and Hostile Audiences.” For purposes of the experiment, I consider the cases collected in this section as a “canonical.” The question for the day is: how does a network approach fare at re-creating the canonical case list?

Let’s start with the raw data. This section of the textbook has three main cases — Chaplinsky v. New Hampshire (1942), Feiner v. New York (1951), and Cohen v. California (1971). It also has eight squib cases — Gooding v. Wilson (1972), Texas v. Johnson (1989), Cantwell v. Connecticut (1940), Terminiello v. Chicago (1949), Edwards v. South Carolina (1963), Cox v. Louisiana (1965), Kunz v. New York (1951) and Forsyth County, Georgia v. Nationalist Movement (1992).

For our first look, consider the two-degree network linking the newest main case (Cohen) to the oldest main case (Chaplinsky). Using a Spaeth projection/genealogy filter, it looks like this:

As usual, blue upward-facing triangles indicate Spaeth-coded “liberal” decisions where the First Amendment “won” over state law/conviction suppressing speech. Red downward-facing triangles indicate Spaeth-coded “conservative” decisions where the First Amendment “lost.” Note that Chaplinsky is green because, like all pre-1948 cases, it does not have a Spaeth code. Since Chaplinsky upheld a conviction for speech (fighting words!), it should considered as a “conservative decision.

This map shows that the two-degree network connecting Cohen to Chaplinsky picks up the third main case in the section (Feiner) as well as three of the squibs (Terminiello, Edwards, and Cox). This is not too bad. The map also reveals that Sullivan & Feldman did not include as squibs four other cases cited by Cohen, which in turn cited Chaplinsky (Winters, Giboney, Roth and Street). Thus, we can say this simple network is both under- and over- inclusive when compared with the canonical list.

Now let’s change up the network a little. Instead of main cases (Chaplinsky and Cohen), let’s use the earliest and latest latest squibs to anchor the network. Using a Spaeth projection without a genealogy filter, this is what the two-degree network linking Forsyth County (1992) to Cantwell (1940) looks like:

 The result here is interesting. Although the Forsyth County->Cantwell two-degree network picks up more cases than the Cohen->Chaplinsky network (16 versus 10), these cases are less relevant to the canonical list. Specifically, this second network only included two additional canonical cases (Gooding and Terminiello). To me, this vindicates Sullivan & Feldman’s choice of main versus squib cases. In other words, Cohen really is more central to the doctrine at issue. The proof is that Cohen cites more cases now deemed as directly relevant.

Finally, let’s go Very Large. This is the three-degree network connecting Forsyth County to Cantwell.

To get a better look, click on the image. Although the image is still pretty small and does not yet permit zooming in (software still in development!), you can hover your mouse over cases to see their names and even link to the underlying opinions.

The three degree network is indeed Very Large — at least from a doctrinal point of view. It contains 109 cases. Given this, it perhaps comes as no surprise that every single one of the canonical cases from Sullivan and Feldman’s list does get swept into the network. All eleven cases plus a whole lot more!

Although I confess that the main impetus for this experiment was to create a pretty picture, I do think the massive over-inclusiveness of this network is independently significant. I have an intuition that the relevant cases in a doctrinal line will almost always exist within three degrees of citation to each other. This seems to back up that intuition. Of course, I will need to do much more work to properly test this hypothesis. Perhaps the “casebook as canonical list” method explored in this post provides a potential framework to conduct a more rigorous investigation.

As always, reader thoughts are welcome!

This week I’ve been mapping out the Supreme Court’s “clear and present danger” doctrine. Post #1 visualized the 50 years from Schenck (1919) to Brandenburg (1969) according to the narrative in Sullivan & Feldman’s leading Con Law textbook. Post #2 examined an apparent gap in that narrative between Dennis (1951) and Brandenburg. Based on a citation network analysis, I hypothesized that the key shift from weak to strong free-speech protection occurred between Communist Party of USA v. Subversive Activities Control Board (1961) and Gibson v. Florida Investigative Commission (1963). Today I wrap up this series with a new map that illustrates the thesis that the Court’s modern era of liberal incitement jurisprudence began with Justice Frankfurter’s departure and replacement by Justice Goldberg.

To set the stage, let’s zoom in on the period in question. Here is a snippet of the Dennis to Brandenburg network from 1960 to 1964.

Note first how the opinions above are proportionally spaced along the X-axis. Unlike the usual evenly-spaced-opinion schematic, this is a genuine timeline view.

The key part of the map is the movement along the line from Scales (1961) to Communist Party (1961) to Gibson (1963). (Smith, the earliest case on the map concerns obscenity not incitement/subversion. I include it simply to help define chronological boundaries.) Oversimplifying just a bit, Scales and Communist Party were both 5-4 decisions where the majority effectively acquiesced to McCarthyism. The majority and dissenting lineups were the same in both cases: [Majority] Frankfurter, Clark, Harlan, Whittaker; [Dissent] Black, Douglas, Warren, Brennan. Notably, Frankfurter wrote the majority opinion in Communist Party. [To see lineups on Oyez, click here for Scales and here for Communist Party].

Then came turnover at the Court. In March 1962, Justice Whittaker was replaced by Justice White. As we’ll see, this change did not turn out to be significant. In August 1962, Justice Frankfurter was replaced by Justice Goldberg. Then the Court handed down Gibson, which effectively pushed back on anti-communist investigations by a state subversive-activities type board. Not coincidentally, Justice Goldberg wrote the Gibson opinion. The new “strong free speech” majority consisted of Warren, Brennan, Goldberg, Black and Douglas. The dissenters were Clark, Harlan, Stewart, and White. Thus we see that White followed the same “weak free speech” line that Whittaker did in Communist Party. [Gibson lineup via Spaeth here].

We can visualize the sequence just described using a timeline map. Click on it to open a full-sized image in a separate window.

The green lines at the bottom of the map indicate when Whittaker and Frankfurter left the Court. To amplify the timeline concept, I also added magenta lines to display important political events of the time.

The map reminds us how intense this period was. Eisenhower left office in 1961 with his famous “military-industrial complex” speech. The Court handed down Scales and Communist Party after President Kennedy took the reins. Then you see the replacement of Whittaker and Frankfurter interspaced with the Bay of Pigs and Cuban Missile crises. After Gibson, you have the March on Washington and Kennedy’s assassination. The times they were a changing!

It probably comes as no surprise to First Amendment scholars that Frankfurter’s departure was a key development in this doctrine. After all, he wrote the majority opinion in Gobitis (1940) which upheld the compulsory saying of the Pledge of Allegiance. Frankfurter then dissented in Barnette (1943), which reversed Gobitis and now stands as another classic statement of modern First Amendment values. Frankfurter was a patriot and a tireless advocate of judicial deference. This served him well often, but not so much on First Amendment questions. At least that is how I see it.

Even if my substantive First Amendment analysis misses the mark, I hope there is some use to the Timeline form explored above. I think visualizing the Court’s relationship to events of the day — placing its doctrinal decisions in political context — is a fruitful endeavor. As always, I’d love to hear what others think. What other visualizations have you seen that put Court doctrine in political context?

This is the second installment in a series charting out the Supreme Court’s famous “clear and present danger” line of decisions. In my first post, I proposed a map of this doctrine hyperlinked to the cases analyzed in Sullivan & Feldman, First Amendment (4th Edition). The resulting picture featured two competing lines: the “bad tendency” opinions versus the “imminent incitement” opinions. While the “bad tendency” line dominated the Court’s majority opinions during the early days, the epic “imminent incitement” dissents penned by Holmes and Brandeis eventually won the day. How did this happen? When did the “imminent incitement” school move from dissent to majority? Today I try to tackle these questions by turning to network theory.

As a starting point, let’s return to the Sullivan & Feldman text. The narrative in the text basically jumps from 1951 (when Dennis upheld a “bad tendency”-esque conviction) to 1969 (when Brandenburg announced the modern incitement test). This is a significant gap and all Sullivan & Feldman offer to bridge it is a brief squib to Bond (1966) as well as the observation that “[b]y the 1960s, the Court had become more protective of speech in connection with the civil rights movement, and the fear of domestic communism had abated politically.” In fairness, Sullivan & Feldman no doubt made the right pedagogical call. Their textbook gives students learning modern doctrine all the context they need. Yet we as scholars can still ask: what happened between Dennis and Brandenburg?

One way to answer this question is to construct a citation network. The map below is an automatically generated 2-degree citation network. This means that the network includes all the cases cited by Brandenburg that in turn cited Dennis.

Note that this is a Spaeth projection. Cases coded by the Supreme Court database as having “conservative” outcomes are red and occupy the lower half of the map; cases coded “liberal” are blue and occupy the upper half. The Y-axis shows the vote for outcome with unanimous judgments occupying the middle.  (For a video explanation about Spaeth projections, see here).

Under the Spaeth rubric, those decisions that uphold free speech rights are “liberal” and those that deny a free speech violation occurred are “conservative.” Looking at this map then, we can see that between Dennis and Brandenburg, the Court handed down 5 liberal/strong-free-speech decisions (Yates, Speiser, Noto, Baggett, and Aptheker) and only 2 conservative/weak-free-speech decisions (Barenblatt and Scales). Using what I call a “genealogy” algorithm, we can then display the citations between cases in the network to highlight the competing lines.

The map above shows cases going both ways after Dennis, but hints at some tension from 1959 through 1964. During this period you have two 5-4 conservative decisions (Barenblatt and Scales) followed by a 7-2 liberal (Bagett) and then a 6-3 liberal (Apthekar). We need to look closer. One way to do this would be to read the opinions. (Readers interested in this approach can click on the map and cases therein to access opinions directly). Another way is to dive deeper into the network. Let’s try that approach.

This third map is the 3-degree citation network connecting Brandenburg to Dennis.  (To open a full-sized map in a separate window, click on the image). This network therefore includes all the 2-degree cases as well as 3-degree cases (i.e., cases cited by 2-degree cases that in turn cite Dennis). I also ran this network through the genealogy algorithm, which cleans up the network considerably.

One pattern emerges particularly sharply from this map. At the bottom, we see a whole series of linked 5-4 conservative decisions extending from 1951 (Garner v. Board of Public Works of Los Angeles) all the way to 1961 (Communist Party v. Subversive Activities Control Board).  Yet after 10 decisions, this line abruptly stops. In 1963, we suddenly see a 5-4 liberal decision that cites Communist Party — Gibson v. Florida Legislative Investigation Commission. The tide has turned. After Gibson, the network consists of exclusively liberal decisions all the way to Brandenburg. This pattern strongly suggests that something — or some things — happened between the Court’s decisions in Communist Party (1961) and Gibson (1963).

The network approach thus helps us shape an inquiry and sharpen our focus. We are able to identify many potentially relevant cases — 33 in the 3-degree network between Dennis and Brandenburg — and then narrow our gaze to a smaller subset based on an observed shift. This seems useful to me. Of course, work remains to be done. Creating the network does not end the inquiry so much as move it along.

In my next post, I will focus in on the key period between 1961 and 1963 identified through this network analysis. Until then, stay tuned!

** Note to First Amendment Researchers ** If you wish to read or examine the opinions in the 3-degree map above, you’ll have to use this link. Due to current technical limitations, this map is a little on the small side. However, when used in conjunction with this full-sized image, you should be able to figure out all you need to get the right links.