Jaded investors have always suspected certain CEOs provide something less than the truth, the whole truth and nothing but the truth when they pick up the phone for their quarterly conference calls with analysts and investors.

Their suspicions are only confirmed -- if they are ever confirmed -- months later when the real story finds the light of day, usually because of regulatory scrutiny or a corporate confession.

But imagine if there was a way for those listening in to predict whether a CEO or other executives on the call were lying.

Because most quarterly earnings webcasts don't include video, using body language to judge whether they are telling the truth is out. But researchers at Stanford University believe you may be able to reach some conclusions based on the types of words and phrases CEOs use during the calls.

David F. Larcker, co-director of the school's Rock Center for Corporate Governance, and Ph.D. candidate Anastasia Zakolyukina used computers to search transcripts of more than 29,000 conference calls. What they were looking for was language that could predict whether the company would restate its earnings sometime in the future.

They found that CEOs who were being deceptive were more likely to use impersonal pronouns (everybody or anybody, for example) rather than refer to themselves. They also used references tied to general knowledge such as "you know" or "everybody would agree."

Moreover, CEOs were more likely to use words expressing extreme positive emotions ("fantastic" rather than "good," for example); used fewer words expressing extremely negative emotions; and were less likely to refer to shareholder value.

Dr. Larcker and Ms. Zakolyukina said similar patterns emerged for CFOs, although they did not use as many extreme words when it came to expressing positive or negative emotions.

The researchers say their method of ferreting out a lying CEO is only 4 percent to 6 percent more accurate than the 50-50 odds you'd get by flipping a coin. But, they say it's a significantly better predictor of whether a company's financials subsequently will be restated than the method employed by researchers who try to predict restatements by examining how a company takes advantage of discretion provided by accounting rules.

"Our results suggest that linguistic features of CEOs and CFOs in conference call[s] ... can be used to identify deceptive financial reporting," the researchers wrote.

Ms. Zakolyukina said if the executives are aware that their company's accounting is deceptive and that causes them stress, that will change the language they use on the call.

There is a growing body of research on what can be learned about Corporate America by the language used by its highest practitioners. A recent study by Penn State management professor Vilmos Misangyi and two colleagues concluded that the more optimistic language new CEOs used in their first letter to shareholders, the more likely analysts were to issue favorable recommendations about the stock.

Some studies are based on how often certain words or types of words are used, while others analyze the tone of the message, the Stanford researchers wrote. They developed their list of words and types of words by reviewing transcripts from 10 conference calls by companies who later restated their financial results.

They looked only at the question-and-answer portion of conference calls, figuring the formal presentations that begin each session are carefully scripted and rehearsed.

Even though rehearsals include practice answering questions analysts and investors are likely to ask, they concluded that the language CEOs and CFOs use in the Q&A portion will be more spontaneous and therefore a better indicator of whether they are telling the truth.

Dr. Larcker and Ms. Zakolyukina caution that deception is very difficult to find and that their findings would benefit from additional research. They said the words they identified may not be the right ones to determine whether CEOs and CFOs are being deceptive. Another potential issue is their assumption that executives know at the time of the call that their accounting has been manipulated, which may not be the case.

In addition, they note, just counting how often certain words are used may not reflect the context they are used in.

One possible subject of additional research would be studying the same CEOs over a long period of time to determine whether they use different language on calls about earnings that will later be restated than they do when discussing earnings that aren't subsequently restated.

Dr. Larcker says he and Ms. Zakolyukina have begun studying whether investors can avoid losses by analyzing a CEO's language. If studying executive speech patterns could predict future earnings restatements with a high degree of certainty, investors could avoid a loss by selling the stock before it moves lower because of the accounting confession.

"We are working on this project right now," Dr. Larcker said.

"Money Q&A" and "Company Town" are featured exclusively at PG+, a members-only web site of the Pittsburgh Post-Gazette. Our introduction to PG+ gives you all the details.