By May of 1980 I had enough hours in speech and related subjects to qualify for a PhD. My oral exams and my dissertation remained.

I needed to form a committee. I think that Steve Alderton¹ was assigned as the head of my committee. I doubt that I chose him. I did choose George Ziegelmueller, the Director of Forensics and also my boss, and Ray Ross, the author of the Speech 100 textbook. Other graduate students assured me that Ross was a soft touch.

I was not worried about the orals. I reviewed a few notes for maybe an hour just before the test began. A psych professor had told us that the average performance on oral exams is horrendous because most students get flustered. The best performance he had seen was by a mediocre student who also hosted a program on the university’s radio station. I figured that my 14+ years of debate experience was more valuable than that. I knew that the trick was to admit it quickly when you didn’t know something. Don’t try to invent an answer. That is, maximize the time spent on what you know by minimizing the time spent on what you don’t know.

There was one difficult question that I knew that I had to answer. Steve asked me whether validity or reliability was more important in a statistical study. I mulled over the question for a few seconds and then chose validity. I hedged my bet by saying that reliability was important, but if your study lacked validity, you did not have anything. I am pretty sure that I gave the right answer.

Anyway, the committee only kept me waiting for about five minutes before they told me that I passed.

The topic of my dissertation was communication in groups. I was most interested in the power of arguments. Before I describe what I proposed to do, I need to talk about a group-communication study on which I worked with Steve earlier in the year.

The data for Steve’s study was collected before I became involved. Forty or fifty people were presented with two different problems that were each described in two or three paragraphs. They were asked to choose between two alternatives in both cases. Their responses were recorded.

Then they discussed both problems as a group. My recollection is that there were fewer than ten groups. Each group turned in a recommendation for each problem. The discussions were recorded on tape. Someone transcribed these onto paper.

Steve and I then categorized some of the utterances in the discussions as arguments. The idea was to use statistical tools on the arguments to determine how powerful they were in producing the results.

For some reason Steve was only interested in one of the two problems. We spent a long time reading the transcripts and marking them up. He had somehow established a minimum level of agreement about what constituted an argument, and he claimed that our two evaluations had met this standard.

The final step was to find the correlation between the arguments and the conclusion. The statistical tools required that the items being counted are independent of one another. Well, most people in the social sciences would consider the groups independent of one another. If not, there would be almost no studies of groups. The individuals could also be considered independent, at least when they were filling out their original forms heard from other participants.

Steve, however, wanted to use a method called “conversational analysis” that someone at his Alma Mater, Indiana University, had advocated. In this method the arguments themselves were used to determine the sample size (always designated by the letter n). If you counted the groups, you would probably need at least one hundred of them to have any chance of getting a statistically significant result. Even if you counted individuals instead of groups, the sample size of this experiment was not very large.

However, if you set n based on the count of the arguments, and dozens of arguments could be identified in each discussion, it would be much easier. To me it seemed clear that the people in the groups were not independent of one another. It is even clearer that the arguments should not be considered independent of one another.

Steve had offered to add my name to the paper in which he summarized the findings. When I told him that I did not want him to do this, he asked me why. I told him that I thought that he was calculating his sample size wrong, and this decision made it much easier to get positive results. He responded that quite a few studies that used conversational analysis had already been published.

Believe me; I thought of a lot of sarcastic ways to explain my reluctance to be involved. I did not let any of them past my teeth, but my face may have betrayed how worthless I thought that this approach was. To say Steve was insulted would be a gross understatement.

What did interest me was the problem that he discarded. I don’t remember all of the details, but it involved a student who was challenging his final grade in a class. The grade was based on four tests. The letter grades for the tests, one of which was an F, were provided, as was the final grade assigned by the teacher. In the text it said that the teacher had not erred in his calculations. I think that the four test grades were B, B, C, and F, and the final grade assigned was a D.

In several groups, one enterprising member calculated the final grade the way that one would calculate a GPA: (4 for A, 3 for B, 2 for C, 1 for D, and 0 for F). By this method, the student had four grades with a total of eight points. The group member argued that the student should have received a C because 8 (3+3+2+0) divided by 4 is 2. In every group in which the argument was made, the group’s decision went in favor of the student. When it was not, the decision went the other way.

It was a perfect argument!

Well, like Pope Urban II’s famous speech that launched the first crusade, it was effective, but I would hardly deem it “perfect”. In the first place, the text of the problem explicitly stated that the instructor had not made this kind of a mistake. Furthermore, the 4-3-2-1 method is never used in grading individual tests because the range for an F is too great. What if the F on one test was, for whatever reason, a 0? If the B’s are 85’s and the C is a 75, the average grade is 61.25, a low D.

I thought that it might be interesting to explore why people in the groups capitulated to what seemed to me a poor argument. However, it was Steve’s data, and I did not have the gall to ask him for it to write a competing paper. As it was, he was very irritated with me already.

For the dissertation committee I also needed to recruit someone from outside the department. I planned to ask the professor who taught the psych class that I had aced and who explained about the orals. However, when I finally got around to asking him, I learned that he was on sabbatical. I really had no choice but to ask the psych department to provide a substitute. I sent the assigned professor a copy of my prospectus, but I had never actually talked to him!

For my study I wanted to use the “shift to risk” studies to which I had been introduced by Prof. Cartwright in the psych department at U-M. At that time at least forty papers had been published that used exactly the same set of twelve problems². The original study had concluded that groups made riskier decisions than individuals. Some later studies found that this shift only occurred on nine of the twelve problems. One problem showed negligible change, and in the other two the group decisions were actually more conservative.

In these studies the answers were always given in terms of a probability of success at which the riskier choice would be desirable. That is, people were asked to assign a number between  one in ten and and ten in ten that represented the lowest chance for success for which they would recommend the riskier alternative. My hypothesis was (1) that each of the twelve problems had a natural set of arguments; (2) not all people are accustomed to making risk-reward decisions based on Bayesian probabilities; (3) these people can be swayed by arguments that they had not considered.

Steve asked the committee to determine whether this study was (1) important enough for a dissertation, and (2) really about speech communication. The psych professor immediately spoke up. He said that he could not address the second issue. However, he said that mine was a very clever approach to an important topic. Despite the fact that there had been a large number of studies to analyze these shifts, no one had ever proposed this mechanism. I was dumbfounded by this unrequested assistance, and all the committee members were very impressed.

Steve insisted that I add conversational analysis to my methodology in order to bring it under the rubric of speech communication. I agreed to do so, but as I was telling the committee this, I said to myself that I would never spend a minute on this study. It now seemed to me like a potential nightmare that might drag on for years.

I also realized that I really did not want to be a professor of speech communication. I loved debate, and, at least at that time, the only way to coach debate was to be on the speech faculty of a school with a debate program. I probably would need to fight for funding for the program, a task that I would not enjoy. Furthermore, because of my background in math and statistics, I felt certain that I would be asked to sit on every committee that evaluated a statistical study. I had heard about and even participated in a few of these, and I had yet to encounter one with which I would want to be associated. Here are a few examples.

1. My first postgraduate class at U-M had been an introduction to graduate studies. In it one of the students asserted that he wanted to go to an Arab country to study their television shows to determine how much they widespread and influential American shows were. He wanted someone to finance him to go to Arab countries, watch television for a few months, and take notes. For a PhD!No, he did not speak Arabic. When I asked if they had something like a TV Guide that he could analyze, he said that that would not be sufficient. He said that he needed to see how many camels there were and stuff like that. I am not joking.
2. At Wayne State I participated in a study in which the experimenter obviously lied about what was happening in other aspects of the experiment. The fact that I figured this out should invalidate his approach. It is absolutely not allowed. If I were on his committee, I would have made him start over with a new design.

3. I also read Juddi Trent’s dissertation. She found that the speeches in Nixon’s 1960 presidential campaign differed significantly in style and content from the speeches in 1968. It was a well-known fact (and one that she acknowledged in the first paragraph of her dissertation) that Nixon wrote the first set, and he employed a team of speech-writers for the second set. Her null hypothesis was that the techniques were the same. She then used statistics to prove that the two sets were not likely to be identical, something she knew for a fact before she started.

Faculty members have three main responsibilities: teaching, publishing, and serving on committees. I had little interest in the subjects I would need to teach and publish, and I would be considered an ogre by all of the graduate students. I decided to do something else with my life.

What I decided to do was to try to help Sue’s fledgling computer software company become more viable. Since Michigan was in the throes of one of its increasingly frequent “auto depressions”, we decided to move back to Connecticut.

---

1. While writing this I discovered that only a few years later Steve Alderton changed careers entirely. He got a law degree and then became (for almost three decades) an official of the federal government, a world traveler, and an artist! His obituary is here.

2. This is called the “Choice Dilemmas Questionnaire” (CDQ) published by Kogan and Wallach in 1964. Here is an example: “Mr B, who has developed a severe heart ailment, has the choice of changing many of his strongest life habits or undergoing a delicate medical operation which may succeed or prove fatal.” Participants are asked to read the statement and then imagine they are advising the main character. They are then asked to indicate the probability of success (from 1 in 10 to 10 in 10) sufficient for them to choose the risky alternative.