Did Reagan Decline?

INTRODUCTION

When Ronald Reagan disclosed (via handwritten letter) that he had Alzheimer’s Disease, it rekindled a lingering debate: did dementia affect Reagan during his time as president? While some analysts claimed President Reagan became “intellectually impaired” over time,1–6 others contended he remained “mentally sound in office.”7–9 The debate persisted because the evidence was selective, subjective, and anecdotal.

Against that backdrop, Berisha et al. stepped in with a clever systematic test. 10 Scouring Ronald Reagan’s 46 formal press conferences for linguistic markers linked with dementia, the researchers reached a striking conclusion: “President Reagan was not diagnosed with AD [Alzheimer’s Disease] until August of 1994, but the results of our analyses suggest that changes in speaking patterns were becoming detectable years prior to clinical diagnosis.”^10,p962

Though Berisha et al. were clear that their study did not (and could not) prove whether Reagan had early-stage dementia as president, the evidence was gripping, which made the inference tempting. The New York Times covered the study in a full story under the headline, “In Reagan’s Speeches, Early Clues to Dementia.” 12 Scholars have cited Berisha et al. more than 100 times, including Eysenck and Keane’s textbook, Cognitive Psychology, which summarizes the contribution: “[Reagan] showed a substantial reduction in the use of unique words during his time as president,” presumably “due to mild cognitive impairment.” 13

This study revisits Berisha et al.’s analysis and results. I begin by using their data and replicating their results, including a brief extension to demonstrate the “outlier” press conference excluded from their analysis does not change the findings.

Next, I critique the research design. I theorize that restricting coding to President Reagan’s first 1,400 words (excluding introductory statements) inadvertently introduced a crucial confound: reporter questions. More to the point, I show that White House reporters increasingly posed “follow-up” questions as Reagan’s tenure wore on, which I theorize produced an illusion of redundant responses.

Recoding the President’s press conference answers to distinct questions – the first question posed, as well as aggregated responses to the first 10 distinct questions – I find Reagan’s unique word usage rate held form throughout his eight years as president. I conclude by considering the implications for Reagan’s legacy and Alzheimer’s research.

Replication

Berisha et al. collected Reagan’s (n = 46) and Bush’s (n = 101) press conference transcripts on the idea that presidents’ mental processing manifests in unscripted responses to independent questions. The researchers thus removed the presidents’ prepared statements and reporters’ (and others’) interjections to produce clean transcripts of presidential responses. For comparability, each press conference corpus was capped at the president’s first 1,400 words. The authors explain: “It is known that lexical statistics are dependent on the length of the analyzed document. As a result, to control for the length of the transcripts, we restricted our analysis to the length of the shortest record in the RR set: 1,400 words.”^{10,pg . 961}

Curated, standardized transcripts in hand, the researchers then identified specific linguistic markers among the presidents’ words: unique words; non-specific nouns and fillers; and low-imageability verbs. The first two markers, i.e., unique words as well as non-specific nouns and fillers, showed significant changes for Reagan but not for Bush. Because unique words strike me more concrete conceptually and more valid empirically (i.e., unlikely to depend on a transcriber’s discretion to include vague irrelevancies like “um” and “ah.”), it is the one I revisit below.

Before jumping into the replication, it is worth mentioning that Berisha et al. used press conference number as a proxy for time. It turns out the correlation between press conference number and Reagan’s days in office is 0.99, so using transcript number is indeed synonymous with time.

Figure 1 displays Reagan’s number of unique words during each press conference, using data provided by the authors. The results replicate.^{10,pg . 962} One point worth mentioning is that Berisha et al. excluded Reagan’s February 18, 1982 press conference as an “outlier,” defined as such because it fell more than two standard deviations from the mean. My instinct is to keep the case included (and will going forward), though it turns out not to matter in this instance. The bivariate correlation between unique words and press conference number is -0.45 with the “outlier” case excluded, -0.47 when it is included.

OPEN IN VIEWER

Fig. 1

Reagan’s Unique Words per 100 in his First 1,400 Words of Response, by Press Conference. Note: Replication of Berisha et al.’s (2015) analysis and result. The figure plots Reagan’s unique word usage (among his first 1400 words in response to reporter questions) across his 46 press conferences. Berisha et al. excluded the February 18, 1982 press conference (shown here as an empty datum) as an “outlier.”

The Qs before As

Modern presidential press conferences generally include two segments: the president gives an introductory statement, then there is a back-and-forth with reporters asking questions and the president giving responses. Empirically, it is easy to separate reporters’ questions from presidents’ answers. Conceptually, though, there is good reason to suspect the questions posed influence the answers given. Actually, it would be remarkable if the two— Qs and As— were uncorrelated.

That caveat has an important implication. To the extent reporters’ questions shape presidents’ replies, the former can confound the latter. The specific concern here arises if the structure of reporters’ questions meaningfully shifted over Reagan’s eight years in office.

As it turns out, researchers have shown that reporters’ behavior did change during this era. White House reporters increasingly adopted an adversarial posture in press conferences, seen partly in journalists’ attempts to pin presidents down via “follow-up” questions. 14 To check if this trend was observable within the Reagan presidency, I counted how many times the term “follow up” (or “follow-up” or “followup”) arose during each press conference. Figure 2 shows the result.

OPEN IN VIEWER

Fig. 2

“Follow-Up” Mentions during Reagan Presidential Press Conferences. Note: Plot of how many times the term “follow up” (or “follow-up” or “followup”) arose during each press conference during Ronald Reagan’s presidency.

It seems White House reporters did ask President Reagan evermore follow-ups. For the first third of press conferences, the average number of “follow-up” mentions was 3, which doubled to 6 for the final third of press conferences (the average for the middle third was 4.5). The overall bivariate correlation between press conference number and follow-up mentions is 0.48.

Inasmuch as reporters asked genuine follow-up questions, e.g., questions seeking clarification or elaboration regarding the president’s response to a preceding question, there is good reason to suspect the President responded in kind, i.e., with follow-up answers that were relatively short and repetitive. In fact, Reagan made that very point at his June 14, 1984 press conference: “You know, you shorten the number of questions we get in with all these follow-ups.” In this way, the rise of follow-up questions could explain President Reagan’s apparent decline in unique word usage, particularly within a fixed range (e.g., his first 1,400 words).

MATERIALS AND METHODS

Seeking a more refined look at Reagan’s press conference rhetoric, I made a key change to the Berisha et al. methodology by homing in on Reagan’s specific responses to distinct questions. To this end, undergraduate research assistants (RAs) scrutinized (randomly assigned) press conferences to identify the first 10 “distinct” questions asked. We defined distinct questions as those that primarily sought new or additional information, which we contrasted to follow-up questions: questions that primarily sought clarification or repetition.

It is worth stressing that our coding criterion was the substantive redundancy of each question vis-à-vis its predecessor, which matters because reporters sometimes invoked the “follow-up” label yet asked a substantively distinct question; at other times they asked a follow-up question without explicitly identifying it as such. In our coding, then, a question’s designation as a follow-up did not depend on the questioner, topic, or whether it was explicitly advertised as a “follow-up.” This definition and distinction proved reliable. My independent coding for a subsample, i.e., the first 10 questions of every 5th press conference (5th, 10th, etc.), matched the RAs’ coding for 86 of 90 (96% ) questions.

The first press conference of Reagan’s second term (February 21, 1985) is illustrative. Below are back-to-back questions: a “distinct” one followed by a “follow-up” question.

As we saw it, the second question primarily restates its predecessor: both essentially ask Reagan why he supports emergency aid to farmers now but not in general. Note that although the reporter used the specific phrase “follow-up,” that was not what rendered it so per our coding. It was the substantive redundancy that met our follow-up threshold.

Now consider a different dyad from the same press conference. Here a reporter asks a distinct question (from its predecessor) before posing one s/he described as a “follow-up” but RAs decided was actually distinct.

The first question is about negotiations with Congress; the second is about fairness. So although both questions relate to farm tax credits, they comprise distinct questions that primarily solicit distinct responses.

RESULTS

First Question and Answer

Naturally, a press conference’s first question is always distinct (as it has no predecessor). Reagan’s responses to first questions, which varied in length but averaged a minute or two, thus afford an especially clean look at his response patterns over time. In fact, that Helen Thomas (United Press International’s bureau chief) was granted the first question at press conferences throughout Reagan’s presidency further pinpoints Reagan’s first-question replies as a critical test.

To reliably estimate Reagan’s unique word usage rate, i.e., unique words per total words, one must measure the numerator (unique words) while holding the denominator (total words) constant.10,15, 10,15 I therefore adopted a sampling strategy to calculate Reagan’s unique words among a fixed number of words. Specifically, I isolated every word of Reagan’s first-question response and then randomly sampled 100 words therein. 1 Identifying the number of unique words among the randomly sampled words thereby provided a reliable estimate Reagan’s unique words per 100 words.

Although Reagan’s single-response “speech data” samples are relatively short, 16 Fig. 3 shows that his unique word rate was remarkably consistent. When answering the first question of each press conference, the President’s average unique word usage rate (per 100) was 70.6 through the first third of press conferences, 69.5 during the middle third, and 69.1 for the last third. A bivariate OLS regression test fails to reject the null hypothesis that Reagan’s unique word rate was unchanged over time (Supplementary Material 1 details the full OLS regression results).

OPEN IN VIEWER

Fig. 3

Reagan’s Unique Words per 100 in Response to the First Question, by Press Conference. Note: Plot of Reagan’s unique words in a random sample of 100 words drawn from his response to the first question at each press conference.

First 10 Distinct Questions and Aggregated Answers

For a broader view, let me turn to President Reagan’s aggregated responses to the first 10 distinct questions posed at each press conference. As before, I normalize each press conference’s word count by extracting Reagan’s answers, isolating each word therein, and then randomly sampling a fixed number of words from each presidential press conference response corpus. 2 Coding how many randomly sampled words were unique thus provides reliable estimates of Reagan’s unique word usage rate during each press conference.

Per Fig. 4, I again find Reagan’s unique word usage rate indicates neither a sizable nor reliable decline.

OPEN IN VIEWER

Fig. 4

Reagan’s Unique Words per 100 in Aggregated Responses to the First 10 Distinct Questions, by Press Conference. Note: Plot of Reagan’s unique words in a random sample of 500 words drawn from his aggregated responses to the first 10 distinct questions at each press conference. I converted to unique words per 100 for comparability.

Though the President’s aggregated replies included a slightly lower proportion of unique words over time, the trendline is substantively small and statistically insignificant. More to the point, a bivariate OLS regression test fails to reject the null hypothesis that Reagan’s unique word rate stayed steady throughout his 8 years and 46 press conferences (Supplementary Material 2 details the full OLS regression results).

DISCUSSION

Ronald Reagan provides Alzheimer’s researchers with an extraordinary test case because the President’s words and deeds were meticulously logged more than a decade before his diagnosis. To understand how and when Alzheimer’s Disease affected Ronald Reagan is to learn as much about the disease as the President.

In that spirit, this study built on a particularly noteworthy finding: linguistic markers in Reagan’s answers to reporters’ questions indicate that he was suffering from dementia during his time as president. Here I sought to re-examine Reagan’s response patterns after accounting for the number and types of questions reporters asked. I thus distilled Reagan’s responses to distinct questions, i.e., the first question, as well as his aggregated responses to the first ten distinct questions, and calculated the President’s unique word usage rate (i.e., unique words per 100 words) at each press conference.

The results refute the thesis that Reagan’s press conference performances indicated early-stage dementia. When answering the first question at each of his 46 formal press conferences, Reagan’s rate of unique words held form throughout his two terms. Likewise, Reagan’s unique word usage rate in his aggregated responses to 10 distinct questions also remained equivalent – conference to conference, year to year. So even if Reagan gave reductive answers to an ever-increasing barrage of “follow-up” questions, his varied replies to distinct questions contradict claims of a sizeable or reliable mental decline.

Beyond the obvious implications for Reagan’s presidency and legacy, this study makes three useful contributions to Alzheimer’s research more generally. First, one easy-to-overlook virtue of transcripts is that they do not just provide data at a specific moment in time (e.g., by day); they also allow for fine-grained analyses (e.g., by question or minute). Pairing Reagan’s responses to first questions with his aggregated responses to 10 questions provided a higher resolution for understanding the nature of Reagan’s rhetorical patterns across different press conferences and within each press conference.

Second, a familiar challenge to lexical analyses is that “speech data” often vary in length, which confounds comparisons between cases. 16 Yet variation of this sort is as much a virtue as a vice, and restricting analyses to the lowest common denominator entails throwing away valuable evidence. Using random sampling, I was able to appraise Reagan’s full, varied responses while still holding the number of words analyzed constant. Clearly, normalizing speech data via random samples is better than via convenience samples (e.g., the first n words), and this technique ought to prove useful going forward.

That gets to the final scholarly take-away. Although Alzheimer’s researchers know well that detecting early-stage dementia in speech data involves considerable uncertainty, Reagan’s example puts a finer point on the implication. Inferring that Reagan said or did less is different than inferring that his mind declined, and scholars should debate openly what specific tests reveal – or not. Reanalyzing Reagan’s unique word usage rate during his presidency’s 46 press conferences, I find the evidence of his decline is dubious, if not doubtful.

AUTHOR CONTRIBUTIONS

Matthew N. Beckmann (Conceptualization; Data curation; Formal analysis; Investigation; Methodology; Project administration; Supervision; Validation; Visualization; Writing – original draft; Writing – review & editing).