Content uploaded by Edward Stetson
Author content
All content in this area was uploaded by Edward Stetson on Feb 09, 2023
Content may be subject to copyright.
Content uploaded by Edward Stetson
Author content
All content in this area was uploaded by Edward Stetson on Feb 09, 2023
Content may be subject to copyright.
Content uploaded by Edward Stetson
Author content
All content in this area was uploaded by Edward Stetson on Feb 04, 2023
Content may be subject to copyright.
TEXAS LAW R EVIEW 1
Submission Texas Law Review Online -draft /8/2023
[pending application]
Response
Texas Writs: It’s not a book, It’s a weapon! A Response to
Tippet and Alexander et alia, Does Lawyering Matter?
Predicting Judicial Decisions...*
Edward Stetson†
Swinging a Pickaxe
In ‘Does Lawyering Matter? Predicting Judicial Decisions from Legal Briefs, and
What That Means for Access to Justice’, Tippet et alia call for: ”a freely
available, computationally enabled citation identification and brief bank
tool, which would extend to all litigants the benefits of good lawyering
and open up access to justice.” Such a thing they call a ‘tool’ even as the
vocabulary of data science seems to obscure its practicality as an
implement for those novices toiling in some field of the common law.
Their focus is on ploughing the loam of employment law, while others are
confined to wielding a pick axe in the calcareous environs of the Texas
criminal courts.
Consider then the legal context of a prisoner beyond appeal, incarcerated,
indigent, and typically without the professional skills of an employment
law litigant. She may have limited access to the resources of the prison
library. She has no right to indigent counsel in swinging a habeas corpus
writ And not infrequently, she may face the aridity of the bars while
pursuing claims of Strickland incompetence and Brady misconduct. The
Tippett tool would indeed be a useful implement and perhaps, in some
cases, her only means to access to justice, for even the best of pled facts
may be overcome by a drought of legal argument.
* Elizabeth Chika Tippett, Alexander et al., Does Lawyering Matter? Predicting Judicial Decisions from
Legal Briefs, and What That Means for Access to Justice, 100 Texas Law Review (2022), (last visited Jan
7, 2023).
† Edward Stetson, Writs Data Project. Legal-ProSe.net, papers: bail reform common law of writs,
TEXAS LAW R EVIEW 2
The Tool augmented as suggested in this response is not of mere
predictive interest. ‘Wat that means for access’ in the fight for justice is
that, ‘It is not just a book. It’s a weapon.’3
This response proposes a sharpening of the Tippet ‘Tool,” The response
charts a reconceptualization of the problem based on practitioner and
theoretical formulations of the common law. Based on this
reconceptualization, It introduces an independently developed
‘Heuristic‘ that could augment the Tool, providing several advantages.,
Its application is illustrated in results from Habeas Corpus writs of the
Texas Court of Criminal Appeals.
The Heuristic has three major advantages that may help the authors
extend their tool’s applicability namely: 1) parsimony in data and focus, 2)
factual contextualization, and 3) recognition of the time dynamics of the
common law. The heuristic may be then characterized as a supplement to
the more formal ‘intelligent’, ‘machine learning’ methods described by the
authors literature review and in Salib, Huang4 .
In Part I the response proposes a justification for our (and Tippett’s)
formalization of precedent as a vector of citations with context. In Part II it
charts a different conceptualization of the data driven decision problem of
the Tool’ solution. Along the way it distinguishes it from the author’s
approach and delineate its advantages and weaknesses. In Part III, it
focuses on the applicability of the model to the dynamics of the common
law, recognizing that a data driven static model is unlikely to be useful to
litigants operating on a changing legal landscape. 5 The appendix provides
Heuristic results for some of the top citation case clusters the method has
identified . In itself, the appendix can serve as an example of Tippett;s
“computationally enabled citation identification ... bank tool.” There we
suggest further research required for its implementation as a practical
application in ‘good lawyering’.
3THE BOOK OF ELI #5 MOVIE CLIP - THE BOOK IS A WEAPON (2010) HD, (2011), (last visited Jan 28,
2023).
4 Peter N. Salib, Artificially Intelligent Class Actions | Texas Law Review, 100 TEXAS LAW REVIEW
(2022), (last visited Jan 7, 2023).,
Zihan Huang et al., Context-Aware Legal Citation Recommendation using Deep Learning, in
PROCEEDINGS OF THE EIGHTEENTH INTERNATIONAL CONFEREN CE ON ARTIFICIAL INTELLIGENCE AND
LAW 79 (2021), (last visited Jan 28, 2023).
5 Of course, a static model’s recommendation would be worse than useless when a past legal rule has
become effectively null. Scientific rules may also change, Valena E. Beety, Changed Science Writs and
State Habeas Relief, 57 HOUS. L. REV. 483 (2019).
TEXAS LAW R EVIEW 3
Dog Law
The Heuristic follows Schauer’s reformulation of the dynamics of the
common law:
Jeremy Bentham, history’s most famous critic of the common law,
notoriously referred to the common law as “dog law”: When your dog does
anything you want to break him of, you wait till he does it, and then beat
him for it. This is the way you make laws for your dog: and this is the way
judges make laws for you and me” …
Bentham’s suspicions notwithstanding … the common law appears
resistant to erroneous results. If a preexisting common-law rule seems to
produce a bad result in a particular case, then it must be because the
preexisting rule is not yet as “pure” as it could and should be and thus
stands in need of modification. And the impetus for this modification will
be a court’s determination that a result is “bad” in light of the full range of
considerations of principle and policy that the court would otherwise use to
evaluate the wisdom of outcomes or decisions, or to make decisions under
circumstances in which there were no rules at all or in which the particular
case was at the fringe and not the center of an existing rule.6
Under this reformulation, a rule is a cluster(vector) of applicable. citation
precedents with authority. It applies because the fact pattern fits the
decision in question. It is further applicable because the legal context fits
the precedents. The citations have authority because they apply and are
not abrogated nor qualified.
When a case following this rule results in a bad outcome it becomes a bad
case that disturbs the court. The disturbance may result in legal change,
that is, the rule is modified to some extent. This may be judicially
controversial, as indicated by concurrences and dissents, or by an
observed increase in citations and frequency. These writs of controversy
signal legal change in the time series of decisions. Most decisions are not
controversial and we often observe few citations as there is little need to
justify a simple application of a stable rule. 7
Both uncontroversial and controversial decisions then have authoritative
citations that apply. We can describe this in terms of data as a cluster or
vector of citations and contextual keywords or key phrases. This is the
bank Tool. Given a factual or legal context,, it answers the question, what
is the set of citations that is salient to the Court in granting relief? Yet the
practitioner avoids the dilemma of Bentham’s dog, kin to that of the
6 FREDERICK SCHAUER, THINKING LIKE A LAWYER: A NEW INTRODUCTION TO LEGAL REASONING
(2009)
7 Public opinion as reflected in pending legislative action can affect legal change, Anita S. Krishnakumar,
The Common Law as Statutory Backdrop, (last visited Feb 4, 2023), often as a” constrained substitute” for
a PDR, exempt from jurisdictional preclusion, Jonathan R. Siegel, Habeas, History, and Hermeneutics, 64
ARIZ. L. REV. 505 (2022). e
TEXAS LAW R EVIEW 4
prisoner facing the machine in Kafka’s Penal Colony. We need not
discover the rule through punishment, dismissal. Instead we learn the rule
from the experience of the long line of dogs that precede us in the
machine./
The Converse Conceptualization
The Tool begins with a conceptualization that is basically the question,
can we train a tool to identify explanatory factors that predict legal
success?
The Tool proposition estimated is: If these Factors, then Success.
The Heuristic proceeds along the converse: If Success, then these
factors.
Both the Tool and Heuristic share citations as explanatory factors The
Tool’s domain is a set of employment law briefs the vast majority of which
are unsuccessful. . The Heuristic’s domain is a set of successful writ
decisions. The tool also has style factors while the Heuristic employs
keyword, often factual, context. The tool may lack predictive power
because it omits factual context. For example, a brief may fail not because
the legal argument was invalid but as a result of a mismatch between the
facts of a case and the arguments asserted. The argument may be valid,
but the rule (a citation cluster) does not apply.
Moreover a shotgun splatter of citations offered in the data of
unsuccessful briefs yields little information on what cluster rule is salient
to the court. Decisions, not lawyer briefs, evidence judicial salience.
Practitioners, if not data scientists, know this. They have internalized a
method of applying rules to context evolved from decades of experience,
some of this wisdom having been bequeathed by their more senior
associates. To explicate, return to the hard landscape of Texas writs
practice. Walk over the practitioners’ paths. As this response intuits, they
did not idle in the pedestrian flats but in the high places of Schauer’s ‘hard
cases’, peaks of high citation frequency, as illustrated below.
The Writs Landscape
In our data set of habeas corpus writ appeals in the Texas Court of
Criminal Appeals, we have identified 801 successful writs from Jan. 2018
to Nov. 2022. For each decision we have parsed citations and keyword
groups and their frequency. We have truncated citations to those 425
occurring more than once, and restricted keywords to those 587 occurring
more than ten times. This defines explanatory dictionaries of unique
citations and keywords ordered by total frequency in the decision domain.
This defines a surface, or two dimensional matrix of decisions (801 rows)
and the 425 columns of citation frequency and 587 columns of keyword
frequency
TEXAS LAW R EVIEW 5
Figure 1-The citation skyline
Looking east over the data landscape from the west boundary with the left
right axis corresponding to timed decision observations, we observe a
skyline of spikes where a citation with a vertical frequency has been
observed (Figure 1). Note that zero (not shown) is the most frequent
occurrence, with single citations the more frequent, with very few citations
with frequency over two. These usually involve a more considered
discussion of the cited by the court.
Looking down at the landscape from the east (in this chart, only 75 court
decisions and 25 citation frequency columns), we observe mostly single
citation occurrences with occasional double cites, and in this subset a
single peak with a frequency over two (Figure 2). The left right axis
corresponds to the Tool citation vector.
We may also note in the entire success court decision domain only a small
set of the citations are used often as indicated by their total column sum
count (Figure 3). Only 40 key citations have a total domain frequency
count greater than 15. 8This is consistent with the Tool authors conclusion
that very few citations are actually predictive of success.9
8 The most common citation, Ex parte Young, 418 SW 2d 824 - Tex: Court of Criminal Appeals 1967,
(last visited Jan 8, 2023). , is authoritative, “Pursuant to the provisions of Article 11.07 of the Texas
Code of Criminal Procedure, the clerk of the trial court transmitted to this Court this application for writ of
habeas corpus”. Tippett describes these types of citations as staple boilerplate. Tippett, page 1175-1176
9 “any given brief’s particular citation frequency vector was only modestly predictive of a summary
judgment win …we also identified a subset of citations that, when present in a brief’s citation frequency
vector, increased the probability of summary judgment win. We include the presence of the top 100 of
these “information gain” citations in Table 3, associated with a near-tripling of the MCC score.
Information gain here refers to the contribution of those particular features, or variables, to the predictive
TEXAS LAW R EVIEW 6
Figure 2 Citation Topography Sample
Figure 3Citation FrequencyHistogram
Successful writs follow a path of collateral attack with well-known citation
justification. Appeals off the beaten path must use more obscure citations
but their success is evidently less likely. The practitioners tread along
these frequency peaks.
Correlation of Citations and Keywords
We have described judicial precedent as a surface of citation and keyword
performance of the model”, Tippett et al., supra note 1.
TEXAS LAW R EVIEW 7
columns and decision rows. Then the questions become: 1) How are these
columns related to each other. In successful writs, does a given citation
often appear together with other citations, 2) Do factual keywords often
appear together, and 3) Does the factual or legal issue keyword imply a
citation ,or a set of citations?
The relationship between explanatory variables is measured here with
correlation, charted in figure 5 for the top 25 citations. Moderate correlation
is common. Higher correlations imply a relationship. We can rank order
cross correlations for a given citation or keyword, and this, as in Tippett and
Alexander, defines a group, or ‘cluster’, of related explanatory variables.
The constant value of one is of course the diagonal of a correlation matrix.
We might add that the authors embed our conceptualization to some extent
in their network cluster analysis. By focusing in that section on citations
present in winning briefs, they have both discarded implicitly the losing
data, as we do, and, shifted from predictive modeling to conditional
expectations given success. A disadvantage here is doing this as a sub-
algorithm of the larger predictive model. It is not parsimonious. Moreover,
are they constructing variables based on success in a predictive model of
success? Is this a kind of circular embedding that will over estimate their
model’s validity?
In any case, are approach independently reaches their conclusion to that
section:
[We] measured the winningness of each brief’s network “neighborhood,”
or cluster of briefs defined by the presence of at least one citation in
common. Here again, the graph suggests that winning briefs share common
citations, and that good lawyering, to some extent, may boil down to the
ability to identify winning citations to precedent
To which we may add, only if the facts fit the rule, as decided by the
Court.
Now for illustrative purposes consider the example of two frequent
habeas attack simple clusters. Cluster 1 identifies three correlated citations
on the legal rule of Ineffective Assistance of Counsel. Cluster 2 identifies
two correlated citations on the duty to provide a notice of appeal. If we
chart the citations stacked frequency we can see that the correlation
captures how these citations often appear together through time
(decisions) as illustrated in Figure 5. The appeals yellows often stack
together as do the IATC blues, as predicted by the computed correlations.
One can validate a computed cluster by inspecting each citation’s content.
10 A weakness of this initial Heuristic approach is that membership in a
10Example of a Simple Correlation Cluster Validation of the cases in Figure 5: We merely compare how
the computed cases are cited, If topical they remain in the cluster.
TEXAS LAW R EVIEW 8
cluster depends on pairwise relationships. Ideally, the Tool’s multivariate
approach might be better, albeit if less intuitive. This example is a
simplification, the Heuristic in general fins a larger set of citations and
keywords that define a rule
Figure 4 Correlation Topography
The Brady case has particularly high correlates in our data set.
Keywords are part of the data landscape. Many of these suggest legal
phrases such as ‘actual innocence’ or ‘false testimony.’ But others relate to
the factual context of the case as in ‘scientific’ evidence disputes or in the
crime of ‘robbery’, Other citations appear and are included in a rule case
cluster. (See Table 1).
Cluster 1-Ineffective Assistance
Strickland v. Washington (1984) quote: To prevail on a claim of ineffective assistance of counsel, a
petitioner must show that (1) his counsel's performance was deficient, and (2) the deficient
performance prejudiced his defense.
Ex parte Saenz quote: In reviewing claims of ineffective assistance of counsel, we must decide
whether Applicant has demonstrated by a preponderance of the evidence that his counsel's
representation fell bel ow an objective standard of reasonableness
Ex parte Morrow quote: Applicant contends that his guilty plea was involuntary due to the
ineffective assistance of his trial counsel
Cluster 2-Notice of Appeal
Ex parte Axel quote: Applicant contends that his trial counsel rendered ineffective assistance by
failing to file a notice of appeal.
Jones v. State quote: The Texas Court of Criminal Appeals has held that a trial counsel's duties do
not end upon sentencing, but rather, include advising a client concerning the right to appeal and
"taking other steps to pursue an appeal
TEXAS LAW R EVIEW 9
Figure 5 Stacked cluster frequency observations
Brady v. Maryland, 373 U.S. 83 (1963) Correlates
exculpatory
0.86
withheld
0.86
EX PARTE CHILDERS (Tex. Crim. App. 2018)
0.86
Kyles v. Whitley, 514 U.S. 419 (1995)
0.86
materials
0.85
violation
0.58
robbery
0.54
elizondo
0.50
Ex parte Weinstein, (Tex. Crim. App. 2014)
0.49
scientific
0.49
falsehood
0.43
testimony
0.42
actual
0.33
innocent
0.32
Table 1 Brady Ranked Correlates
Dog Dynamics
In an earlier working paper first exploring writs data, we focused on legal
change in habeas corpus related to improper sentencing enhancement.11
We detailed the progression of a rule (a cluster of citations) through time.
We showed, following Schauer, how new rules are a response to a ‘bad
11 Edward Stetson, The Story of Judge Yeary and the Forty Writs, 1 WRITS DATA PROJECT TEXAS,
.
TEXAS LAW R EVIEW 10
case.’ They may abrogate or restrict, embellish or expand, and in this
sequence, embroider. By embroider we mean expand or restrict past
authorities’ application to specific qualifying exemptions or entitlements.
In that study, originally, Garcia/Ramirez was controlling with regard to
improper sentence enhancement upon appeal. Then under Clay a new
entitlement is allowed in a writ of habeas corpus. Subsequently, Pue and
Westerman clarify this decision with regard to retroactivity, latches delay,
and the finality of a conviction under Texas law (dynamics described in
Figure 6), the green shadow shows a shift in authority weight through
time (depending on the context). This is somewhat of a simplification of
course. The opinions, concurrences, and dissents address a variety of legal
contexts in which the rule may or may not apply. It is these ‘Dog Law’
dynamics which require further augmentation of the Tippett Tool, perhaps
in a regime-change specification.12
Figure 6
So, in this example, legal change may result in innovation with subsequent
clarification or redaction. Although we could formalize this further, it
suffices to say that the Tibbett conceptualization does not perhaps
sufficiently recognize the time dependence of authorities, and the
dependence of this authoritativeness on the factual and legal context. Our
suggested augmentation of the MITRE13 methodology is to make this
explicit, as we do in this heuristic. 14
12 Philipp Singer et al., Detecting Memory and Structure in Human Navigation Patterns Using Markov
Chain Models of Varying Order, 9 PLOS ONE e102070 (2014)., for a Hamilton specification see Richard
Breen, Kristian Bernt Karlson & Anders Holm, Interpreting and Understanding Logits, Probits, and Other
Nonlinear Probability Models, 44 ANNUAL REVIEW OF SOCIOLOGY 39 (2018).
13 The MITRE Corporation (authors Branting, Morawski, Balhana, Pfeifer, and Bayer), video
14 The conditional expectation becomes time index as in a Markov chain with memory, as above,
(() |= 1, =( 1)) . Intuitively, a case presents with a given factual and legal context. The
obvious optimal citation vector is the previous vector used in a successful attack with the same or similar
context.In our opinion, a time series specification would be overly specific, one would have to specify
memory lags,. especially at the state wide level with a limited sample size, see Singer et al., supra note
TEXAS LAW R EVIEW 11
Legal change is highlighted in our sample by decision citation vector
length and frequency magnitude often with concurrences and dissent. As
in the court data, cases with multiple opinions have a higher weight. Both
Tippett’s tool and the Heuristic need to make data weighting more explicit
in the model, perhaps by weighting more highly cited cases or keywords
greater than those obscure cases with low citation frequency? This is
implicit in coefficients of a predictive model as in the Tool, suggesting that
a multi-stage model learning might begin with a Heuristic identification of
clusters to be followed with weight estimation on the category of cluster
membership.
Finally, recognition of dog law dynamics includes the implication that the
book of citations and keywords is a weapon, an arrow in the quiver of
good lawyering. But more than that, there is some malleability in how the
tool is sharpened and fit. We see this in Schauer’s bad case example of
improper enhancement The fact pattern and its resolution in the courts
according to the rule was incongruous with the court’s perception of
justice. Consequently, it broadened Mizell’s applicability. A well pled writ
may be more effective than a well-researched one.
A judge can make the book say what she wants it to say.
REDRIDGE
And what if this book don't work?
What if it don't say what you want it to say?
CARNEGIE
Oh, it'll say what I want, I can promise you that. Because I'm going to rewrite
it. I'll keep the parts that work for me and make the rest whatever I need it
to be15.
Next Steps
This response suggests an augmentation of the Tippet tool with
a Heuristic that may sharpen its effectiveness as an implement
improving access to justice.
Advantages
Parsimony in data-we need not consider legal brief failures but
merely court decisions given success
• Parsimony in focus-Of the possible thousands of citations used in
legal Texas Writs briefs we find that less than one hundred are
11.
15 Gary Whitta, The Book of Eli, (2007 , (last visited Jan 28, 2023).
TEXAS LAW R EVIEW 12
demonstrably important in successful Texas writ court decisions.
Moreover, far fewer cover the range of possible collateral attacks.
• Ease of Implementation-It is likely that all court decisions of every
state are available online in pdf or html format, and then can be
parsed for citations and keywords without reliance on any private
data banks, commercial or academic; It can therefore be generalized
to all jurisdictions and subject areas.
• Contextualization- rather than using formal clustering or network
methods, we can simply identify citation and keyword cluster
groups using intuitive rank ordered correlations truncated by
reference to frequency count. This heuristic is then validated by
inspection of legal and factual keyword content, which then points
to additional clusters (rule and factual contextualization).
•
Doubts
The heuristic described here identifies cases that are important along given
dimensions. Five of these dimension clusters are detailed in the appendix
But being careful to remove data overload means we have to truncate the
information. This in turn requires some fairly arbitrary thresholds of
salience. Consider two cases that share only a single citation that is merely
procedural. Even though these cases would have correlation one, they
may have nothing to do with each other. Or consider a case that has a high
correlation with a target dimension but it is very lightly cited by
subsequent cases. A different case with broader coverage but lower
correlation may be more salient. Good lawyering means here using human
judgement in deciding which cases and citations are extraneous.
Secondly, clusters share case citations. Dimensions overlap. cognizability
restrictions require petitioners to pursue a multivariate collateral attack.
See Figure 7. For example, an involuntary plea may be to an improper
enhancement (Garcia is an intersection).
Westerman is important along three dimensions, “Applicant argues that the
repeat-felon enhancement was improper, that trial counsel was ineffective
for failing to investigate it, and that his guilty plea was involuntary.”
Similarly Dotson intersects IATC and Enhancement, Salim Involuntary
pleas and IARC, Mallet new evidence and enhancement. Along with
Westerman, Salcedo intersects three cluster dimension sets, as do arguably
others
TEXAS LAW R EVIEW 13
Figure 7
The Court has recognized some of this as tactical. Issues not raised in the
trial court or upon appeal cannot be cognizable without resort to the
Bradys , IATC, or another avenue of collateral attack. . 16 Again, the
multivariate nature of the landscape argues more for the Tippett approach
than the Heuristic. Although there is a remedy perhaps in refining the
keyword context to extended into coded key phrases. Clusters could then
be conditioned in membership in multiple citation dimension sets. .17 18
16 IATC now through latches has an effective statute of limitations where counsel may suppress evidence
of attorney error, ”Witnesses critical to the substantiation of a defendant’s claims may be impossible to
contact, or corroborative evidence may be tainted, lost, or destroyed”, Alejandra S. Alvarez, Habeas
Mentem: Revisiting Sufficiency-of-Counsel Standards in Post-AEDPA Habeas Corpus Proceedings, 71
FLA. L. REV. 1481 (2019). If the Bar will not enforce retention of records by the trial courts and by
counsel, IATC issues may spill over into Mandamus and Civil actions? Judicial Hot Potato: An Analysis
of Bifurcated Courts of Last Resort in Texas and Oklahoma, 12 TENNESSEE JOURNAL OF LAW AND
POLICY (2018), Edward Stetson, Is Poverty Immutable for the County Judge | Reference Case Study, writs
data project, LEGAL-PROSE.NET.
17Ex Parte Garcia, 624 SW 3d 921 (2021), Ex Parte Dotson, (2Tex: Court of Criminal Appeals 2022),, Ex
parte Salim, 595 SW 3d 844 (Tex: Court of Appeals, 2nd Dist. 2020020), Ex Parte Mallet, (Tex: Court of
Criminal Appeals ,2020), Ex Parte Westerman, Order, (Tex: Court of Criminal Appeals, 2018), Ex parte
Westerman, Yeary Dissent, 570 SW 3d 731 (Tex: Court of Criminal Appeals, 2019), Ex parte Saucedo,
concurrence, 576 SW 3d 712 (Tex: Court of Criminal Appeals,2019),
18 On innocence, see, In re Lester, 602 SW 3d 469 (Tex: Court of Criminal Appeals,2020),. And Ex parte
Tuley, 109 SW 3d 388 (Tex: Court of Criminal Appeals,2002),., and lessor included,
“Mable did not explain any of this. Consistent with Wilson, we should not have simply rejected Mable's
claim of actual innocence and then proceeded to address a due-process defect that Mable never
complained about. Instead, we should have construed Mable's claim of actual innocence as a claim that he
was "guilty only of" a lesser-included offense or that he was factually "ineligible for" the punishment he
received. Under either of these theories, Mable's claim was self-evidently meritorious., Keasler, Ex parte
Saucedo, concurrence, 576 SW 3d 712 (Tex: Court of Criminal Appeals 2019),
TEXAS LAW R EVIEW 14
Intuition
This response charts a survey of the landscape of Texas Writ opinions and
their relevant citations and keywords. It has illustrated the histogram
density of citations and their correlation structure. The surface of citations
shows peak intensities related to judicial controversy and legal change.
The skyline of citation peaks is consistent with these densities and cluster
element covariation. The landscape survey motivates a citation cluster
identification methodology heuristic that simply finds correlated cites and
keywords and ranks them according to coverage in the data and
correlation. Validation is easily done by inspection. The appendix gives
results of the Heuristic identifying the top clusters for pursuit of Habeas
Corpus relief in the Texas Criminal Court of Appeals. This in itself is a
useful reference for scholars and practitioners alike in the law of Texas
writs. Our cluster results are validated by the wisdom of experienced
Habeas attorneys19, namely petitioners should focus on PDRs and appeals
lost, Strickland Ineffective Assistance of Trial Counsel (IATC), Brady
evidence and Mooney false testimony violations, Brady involuntary pleas,
and constitutionality of statutes.
Future Research
In our initial exploration we have identified many citation and keyword
clusters. There should be a methodology to remove intersections of sets
(redundant information overlap).
In analytic footnotes, we suggest refinement of the Tippett tool with
specifications that recognize conditional expectations with the addition of
dynamics and regime change. Identification of Writs of Controversy
where legal change occurs may be used to identify opportunities for novel
legal argument.
Finally, a mobile application that returns relevant citation clusters to users
based upon a legal or factual inquiry should not be difficult to develop
and deploy. This would be the computationally based legal research tool
suggested by Tippett et alia.
Practitioner Appendices
As an aid to Pro Se litigants and Texas Habeas novices, we have provided
the forty most important explanatory factors (linked citations) found by our
Heuristic. We have also given five key clusters. Each cluster has ten correlates
with breadth (non-trivial citation) and up to ten citing case observations. We hope
that this appendix in itself, absent the suggestions above, is a standalone useful
information portal reference for attorneys and their clients.
19 Thank you for scholarship , leadership, and inspiration: Michael Falkenberg, Article 11.07 Habeas
Corpus, (2022),Carmen Roe. (2022). Advanced Appellate Tips. UT 2022, Carmen Roe Law Firm, PLLC -
Houston, TX; Jennifer Laurin UT CLE, research support Audrey Banda, TX CCA
TEXAS LAW R EVIEW 15
Appendix -Top Writ Case Citation Clusters
Top 40 Citations
1, 364, Ex parte Young, 418 S.W.2d 824, 826 (Tex.
Crim. App. 1967)
11.07 authority
11, 52, Brady v. Maryland, 373 U.S. 83 (1963)
Brady suppression
2, 123, Ex parte Wilson, 956 S.W.2d 25 (Tex. Crim.
App. 1997)
Right to file a PDR
12, 46, Brady v. United States, 397 U.S. 742, 757
(1970)
Brady involuntary pleas
3, 88, Ex parte Axel, 757 S.W.2d 369 (Tex. Crim.
App. 1988)
Right to direct appeal
13, 45, Ex parte Torres, 943 S.W.2d 469 (Tex. Crim.
App. 1997)
HC denial, dismissal merits
4, 84, Jones v. State, 98 S.W.3d 700, 703 (Tex. Crim.
App. 2003)
Right to appeal
14, 45, Dulin v. State, ___ S.W.3d ___, Nos. PD-
0856-19 & PD-1857-19, 2021 WL 1202400 (Tex. Crim.
App. Mar. 31, 2021),
133.103 (b) and (d) facial
unconstitutionality
5, 82, Strickland v. Washington, 466 U.S. 668, 687
(1984)
Effective counsel IATC
15, 43, Ex parte Rodriguez, 334 S.W.2d 294, 294
(Tex. Crim. App. 1960),
trial court remand
6, 75, Ex parte Riley, 193 S.W.3d 900 (Tex. Crim.
App. 2006)
Right to file a PDR
16, 38, Ex parte Mable, 443 S.W.3d 129 (Tex. Crim.
App. 2014)
pleas voluntary, intelligent
7, 69, Ex parte Elizondo, 947 S.W.2d 202 (Tex. Crim.
App. 1996),
New evidence, actual innocence
17, 38, Ex parte Palmberg, 491 S.W.3d 804, 808
(Tex. Crim. App. 2016)
pleas misapprehensions
8, 60, Ex parte Crow, 180 S.W.3d 135 (Tex. Crim.
App. 2005)
PDR IATC
18, 35, Ex parte Rich, 194 S.W.3d 508, 513 (Tex.
Crim. App. 2006),
improper enhancement
9, 59, Ex parte Lo, 424 S.W.3d 10 (Tex. Crim. App.
2013)
Constitutionality
19, 35, State v. Wilson, 324 S.W.3d 595, 598 (Tex.
Crim. App. 2010)
involuntary plea, Brady, actual
innocence
10, 52, Ex parte Fournier, 473 S.W.3d 789, 800-805
(Tex. Crim. App. 2015) (Yeary, J., dissenting)
33.021(b)
20, 34, Ex parte Brown, 205 S.W.3d 538, 544 (Tex.
Crim. App. 2006)
newly discovered evidence and actual
innocence
TEXAS LAW R EVIEW 16
21, 33, Ex parte Tuley, 109 S.W.3d 388, 392 (Tex.
Crim. App. 2002)
new evidence plea, innocence
Herrera Schlup (Elizondo)
31, 16, Hill v. State, 633 S.W.2d 520 (Tex. Crim.
App. 1982)
contemporaneous trial objections
improper enhancement
22, 28, Ex parte Parrott, 396 S.W.3d 531, 536-37
(Tex. Crim. App. 2013)
harm in improper enhancement
32, 16, Ex parte Smith, 444 S.W.3d 661, 667 (Tex.
Crim. App. 2014)
plea IATC deportation, latches
23, 26, Ex parte Broussard, 517 S.W.3d 814, 816
(Tex. Crim. App. 2017)
plea, misapprehension, Palmberg,
Mable, Ruiz
33, 16, Atkins v. Virginia, 536 U.S. 304 (2002)
death penalty mental disability
24, 22, Ex parte Patterson, 993 S.W.2d 114, 115 (Tex.
Crim. App. 1999)
plea, IATC improper
enhancement
34, 15, Ex parte Carter, 521 S.W.3d 344, 347-48
(Tex. Crim. App. 2017)
improper cumulation error forfeited in
direct appeal
25, 22, Ex parte Huerta, 692 S.W.2d 681 (Tex. Crim.
App. 1985)
plea, sentencing, concurrence
35, 15, Jackson v. Virginia Jackson v Virginia, 443
U.S. 307, 316, 99 S.Ct. 2781, 61 L.Ed.2d 560 (1979)
evidence sufficiency
26, 22, Ex parte Saucedo, 576 S.W.3d 712, 714-20
(Tex. Crim. App. 2019) (Keasler, J., concurring, joined
by Hervey and Yeary, JJ.)
plea, Wilson guilty only of a
lesser-included Elizondo claim
36, 15, Ex parte Chance, 439 S.W.3d 918 (Tex. Crim.
App. 2014)
unconstitutionality, void statute, appeal
27, 20, Hill v. Lockhart, 474 U.S. 52, 59, 106 S.Ct.
366, 370, 88 L.Ed. 2d 203 (1985)
plea involuntary Strickland test
37, 15, Kyles v. Whitley, 514 U.S. 419 (1995)
Brady evidence materiality
28, 20, Ex parte Reed, 271 S.W.3d 698, 727 (Tex.
Crim. App. 2008)
ultimate fact finder
38, 15, Ex parte Pue, 552 S.W.3d 226, 243 (Tex.
Crim. App. 2018) (Yeary, J., dissenting)
improper enhancement, finality,
excessive sentence
29, 18, Ex parte Perez, 398 S.W.3d 206, 213 (Tex.
Crim. App. 2013)
common law 5 years latches
39, 15, Anders v. California, 386 U.S. 738, 87 S.Ct.
1396, 18 L.Ed.2d 493 (1967)
frivolous appeal withdrawal
30, 17, Mizell v. State, 119 S.W.3d 804, 806 (Tex.
Crim. App. 2003)
illegal sentence
40, 15, Carpenter v. United States, ___ U.S. ___, 138
S. Ct. 2206, 201 L.Ed.2d 507 (2018)
cell-site location 4th Amendment
TEXAS LAW R EVIEW 17
Five Important Case Clusters-Correlates and Citing Observations with Breadth
5, 82, Strickland v. Washington, 466 U.S. 668, 687 (1984)
Ex parte Rodriguez, 334 S.W.2d 294, 294 (Tex. Crim. App. 1960), (43,
0.43)
Ex parte Westerman, 570 SW 3d 731 (Tex. Crim. App.
2019) (4, 1)
Ex parte Patterson, 993 S.W.2d 114, 115 (Tex. Crim. App. 1999) (22,
0.35)
EX PARTE REED (Tex. Crim. App. 2022) (2, 1)
Ex parte Reed, 271 S.W.3d 698, 727 (Tex. Crim. App. 2008) (20, 0.3)
EX PARTE AHMED (Tex. Crim. App. 2018) (1, 1)
Penry v. Lynaugh, 492 U.S. 302, 328 (1989) (12, 0.35)
Ex parte Garza, 620 SW 3d 801 (Tex. Crim. App. 2021)
(1, 8)
Ex parte Weinstein, 421 S.W.3d 656 (Tex. Crim. App. 2014) (11, 0.19)
EX PARTE KELLEY (Tex. Crim. App. 2019) (1, 1)
Ex parte Maldonado, 688 S.W.2d 114, 116 (Tex. Crim. App. 1985)
(9, 0.44)
EX PARTE KOLHOFF (Tex. Crim. App. 2020) (1, 1)
Wiggins v. Smith, 539 U.S. 510, 123 S.Ct. 2527, 156 L.Ed.2d 471
(2003) (9, 0.72)
EX PARTE VAUGHN (Tex. Crim. App. 2020) (1, 1)
Ex parte Gonzales, 204 S.W.3d 391, 394 (Tex. Crim. App. 2006) (7,
0.72)
EX PARTE SHIFLETT (Tex. Crim. App. 2020) (1, 1)
Ex parte Morrow, 952 S.W.2d 530, 534 (Tex. Crim. App. 1997) (6,
0.17)
Ex parte McEwen, 602 SW 3d 586 (Tex. Crim. App.
2020) (1, 1)
Ex parte Nelson, 137 S.W.3d 666 (Tex. Crim. App. 2004), (5, 0.26)
EX PARTE BALL (Tex. Crim. App. 2020) (1, 1)
7, 69, Ex parte Elizondo, 947 S.W.2d 202 (Tex. Crim. App. 1996),
Brady v. United States, 397 U.S. 742, 757 (1970) (46, 0.35)
Ex parte Saucedo, 576 SW 3d 712 (Tex. Crim. App.
2019) (22, 3)
State v. Wilson, 324 S.W.3d 595, 598 (Tex. Crim. App. 2010) (35,
0.51)
Ex parte Warfield, 618 SW 3d 69 (Tex. Crim. App. 2021)
(12, 1)
Ex parte Brown, 205 S.W.3d 538, 544 (Tex. Crim. App. 2006) (34,
0.45)
Ex parte Kussmaul, 548 SW 3d 606 (Tex. Crim. App.
2018) (10, 1)
Ex parte Tuley, 109 S.W.3d 388, 392 (Tex. Crim. App. 2002) (33,
0.54)
Ex parte Chaney, 563 SW 3d 239 (Tex. Crim. App. 2018)
(10, 1)
Ex parte Broussard, 517 S.W.3d 814, 816 (Tex. Crim. App. 2017)
(26, 0.42)
Ex parte Mallet, 602 SW 3d 922 (Tex. Crim. App. 2020)
(3, 1)
Ex Parte Miles, 359 S.W.3d 647, 671 (Tex. Crim. App. 2012) (13,
0.36)
EX PARTE ALONZO (Tex. Crim. App. 2019) (3, 1)
Ex parte Cacy, 543 S.W.3d 802 (Tex. Crim. App. 2016) (Yeary, J.,
concurring) (12, 0.38)
EX PARTE GARCIA (Tex. Crim. App. 2021) (2, 1)
Ex parte Kussmaul, 548 S.W.3d 606, 636-37 (Tex. Crim. App. 2018)
(10, 0.41)
Ex parte Grant, 622 SW 3d 392 (Tex. Crim. App. 2021) (2,
1)
McMann v. Richardson, 397 U.S. 759, 770, 90 S.Ct. 1441, 25 L.Ed.2d
763 (1970) (7, 0.41)
EX PARTE MALLET (Tex. Crim. App. 2020) (2, 1)
Ex parte Reyes, 474 S.W.3d 677, 681 (Tex. Crim. App. 2015) (6, 0.41)
EX PARTE GARCIA, 624 SW 3d 921 (Tex. Crim. App.
2021) (2, 1)
11, 52, Brady v. Maryland, 373 U.S. 83 (1963)
Ex parte Palmberg, 491 S.W.3d 804, 808 (Tex. Crim. App. 2016) (38,
0.34)
Ex parte Chaney, 563 SW 3d 239 (Tex. Crim. App. 2018)
(10, 1)
Kyles v. Whitley, 514 U.S. 419 (1995) (15, 0.39)
Ex parte Mallet, 602 SW 3d 922 (Tex. Crim. App. 2020) (3,
1)
Harm v. State, 183 S.W.3d 403, 408 (Tex. Crim. App. 2006) (14,
0.38)
EX PARTE NICHOLSON, 634 SW 3d 743 (Tex. Crim.
App. 2021) (3, 1)
United States v. Bagley, 473 U.S. 667, 676, 105 S.Ct. 3375, 87
L.Ed.2d 481 (1985) (11, 0.29)
EX PARTE CHILDERS (Tex. Crim. App. 2018) (3, 1)
TEXAS LAW R EVIEW 18
Ex parte Harleston, 431 S.W.3d 67, 89 (Tex. Crim. App. 2014) (8,
0.31)
EX PARTE NICHOLSON (Tex. Crim. App. 2021) (2, 1)
Estrada v. State, 313 S.W.3d 274, 281 n.3 (Tex. Crim. App. 2010) (7,
0.29)
EX PARTE MALLET (Tex. Crim. App. 2020) (2, 1)
Saldano v. State, 70 S.W.3d 873, 884 (Tex. Crim. App. 2002) (7,
0.29)
EX PARTE COLONE (Tex. Crim. App. 2022) (1, 2)
Ex parte Chabot, 300 S.W.3d 768 (Tex. Crim. App. 2009) (7, 0.29)
EX PARTE YEARLING (Tex. Crim. App. 2022) (1, 2)
Ex parte Mayhugh, 512 S.W.3d 285, 298 (Tex. Crim. App. 2016) (6,
0.31)
EX PARTE AUSTIN (Tex. Crim. App. 2020) (1, 1)
Britain v. State, 412 S.W.3d 518, 520 (Tex. Crim. App 2013) (5,
0.35)
EX PARTE JAILE (Tex. Crim. App. 2019) (1, 1)
12, 46, Brady v. United States, 397 U.S. 742, 757 (1970)
Ex parte Elizondo, 947 S.W.2d 202 (Tex. Crim. App. 1996), (69,
0.35)
Ex parte Saucedo, 576 SW 3d 712 (Tex. Crim. App.
2019) (22, 4)
Ex parte Mable, 443 S.W.3d 129 (Tex. Crim. App. 2014) (38, 0.38)
EX PARTE Mackley (Tex. Crim. App. 2020) (2, 1)
Ex parte Palmberg, 491 S.W.3d 804, 808 (Tex. Crim. App. 2016)
(38, 0.47)
EX PARTE GARCIA (Tex. Crim. App. 2021) (2, 1)
State v. Wilson, 324 S.W.3d 595, 598 (Tex. Crim. App. 2010) (35,
0.26)
EX PARTE REYNA (Tex. Crim. App. 2021) (2, 1)
Ex parte Broussard, 517 S.W.3d 814, 816 (Tex. Crim. App. 2017)
(26, 0.39)
EX PARTE JOSEPH (Tex. Crim. App. 2019) (1, 1)
Ex parte Kimes, 872 S.W.2d 700, 703 (Tex. Crim. App. 1993) (10,
0.2)
EX PARTE THOMPSON (Tex. Crim. App. 2019) (1, 1)
Ex parte Barnaby, 475 S.W.3d 316, 325 (Tex. Crim. App. 2015) (8,
0.32)
EX PARTE DIXON (Tex. Crim. App. 2020) (1, 1)
Ex parte Moussazadeh, 361 S.W.3d 684, 690 (Tex. Crim. App. 2012)
(7, 0.34)
EX PARTE HICKS, 640 SW 3d 232 (Tex. Crim. App.
2022) (1, 3)
McMann v. Richardson, 397 U.S. 759, 770, 90 S.Ct. 1441, 25 L.Ed.2d
763 (1970) (7, 0.55)
EX PARTE CAPE (Tex. Crim. App. 2021) (1, 1)
Ex parte Reyes, 474 S.W.3d 677, 681 (Tex. Crim. App. 2015) (6,
0.26)
EX PARTE WESTERMAN, 592 SW 3d 441 (Tex. Crim.
App. 2019) (1, 1)
18, 35, Ex parte Rich, 194 S.W.3d 508, 513 (Tex. Crim. App. 2006),
Mizell v. State, 119 S.W.3d 804, 806 (Tex. Crim. App. 2003) (17,
0.61)
Ex parte Saucedo, 576 SW 3d 712 (Tex. Crim. App. 2019)
(22, 1)
Hill v. State, 633 S.W.2d 520 (Tex. Crim. App. 1982) (16, 0.62)
Ex parte Pue, 552 SW 3d 226 (Tex. Crim. App. 2018) (15,
7)
Ex parte Clay, 539 S.W.3d at 289 (Yeary, J., dissenting) (9, 0.62)
Ex parte Clay, 539 SW 3d 285 (Tex. Crim. App. 2018) (9,
2)
Ex parte White, 659 S.W.2d 434, 435-36 (Tex. Crim. App. 1983)
(7, 0.7)
Ex parte Hill, 632 SW 3d 547 (Tex. Crim. App. 2021) (4, 1)
Ex parte Cashman, 671 S.W.2d 510, 512 (Tex. Crim. App. 1984) (7,
0.54)
EX PARTE PHILLIPS (Tex. Crim. App. 2022) (2, 1)
Ex parte Townsend, 137 S.W.3d 79 (Tex. Crim. App. 2004), (6,
0.54)
EX PARTE PRINGLER (Tex. Crim. App. 2020) (1, 1)
Ex parte Nivens, 619 S.W.2d 184, 185 (Tex. Crim. App. 1981) (6,
0.76)
EX PARTE RENDON (Tex. Crim. App. 2021) (1, 1)
Ex parte Perales, 215 S.W.3d 418, 419 (Tex. Crim. App. 2007) (5,
0.56)
EX PARTE COKER (Tex. Crim. App. 2018) (1, 1)
Ex parte Langley, 833 S.W.2d 141 (Tex. Crim. App. 1992) (5, 0.8)
EX PARTE ORTIZ (Tex. Crim. App. 2021) (1, 1)
Ex parte Nelson, 137 S.W.3d 666 (Tex. Crim. App. 2004), (5,
0.82)
EX PARTE SKIPPER (Tex. Crim. App. 2021) (1, 1)
