IRVING R. KAUFMAN, Circuit Judge:

It is the task of trial judges to separate factual wheat from evidentiary chaff, and appellate courts must accord great deference to these determinations. The temptation to displace the trial court’s judgments with our own is often strong, but the integrity of the decision-making process requires that we do so only in cases of clear error. Fed.R.Civ.P. 52(a). The same policies apply in the administrative context when decisions of the finder of fact are brought under review.

In this case, the National Labor Relations Board appears to have disregarded this principle in reversing credibility determinations made by an administrative law judge. The Board found that the petitioner failed to show he was subjected to retaliatory treatment by his employer. We reverse the Board’s decision, which we find unsupported by substantial evidence. Subsequent to issuing its order here, however, the Board overruled a principle of law on which this case turns. We must therefore remand to the Board for a determination whether its new interpretation of the law should be applied retroactively so as to bar the relief originally ordered by the ALJ.

Many of the facts relevant to this petition are vigorously disputed by the parties. We begin with those that are not. Petitioner, Robert T. Ewing, is a piledriver and a member of the Piledrivers, Dock Builders, Trestle, Crib Breakwater Builders, Local 1978, AFL-CIO (the “Union”). He has worked on numerous construction projects in the Buffalo, New York area. The construction business is by nature cyclical, and piledrivers experience frequent layoffs and recalls as activity in the industry fluctuates in relation to seasonal change and varying economic conditions. Consistent with this pattern, Ewing worked off and on for the construction firm Herbert F. Darling, Inc., from 1967 through early 1980.

Nineteen hundred and eighty, however, was a banner year for construction in west *1119 ern New York and, beginning in March, Ewing was consistently employed by Darling as part of a five-man team of pile-drivers working on a rapid transit project in Buffalo. Ewing was referred to the job by the customary means: a Darling supervisor contacted Union business manager William Burke, and asked for the requisite number of piledrivers. The company sometimes requested workers by name; on other occasions, it simply stated the number needed. Ewing’s team worked through the summer and fall, until all five men were laid off on December 3, with the promise of new work in about a month. On December 19 all members of the crew except Ewing were recalled, and continued to work steadily into the following year. Ewing was not recalled until April 27, 1981, and worked only a week before being laid off again. He was recalled for three brief periods in the ensuing months, and finally, in December, was restored to Darling’s regular employ.

Petitioner’s and respondent’s explanations of the .latter events are in sharp conflict. Ewing testified that when he was not recalled after about a month, he called Burke and told him he was going to the rapid transit job site to find out what was happening. At the site he encountered Richard Radel, project manager for Darling, who told him he expected to have more work in a week or two. Ewing remained in laid off status, however, and sometime later communicated with Burke again. At this point, Burke testified, he told Ewing that he heard that Darling officials believed Ewing had reported the company to the Occupational Safety and Health Administration (“OSHA”), resulting in an-inspection of the rapid transit site in October 1980 which uncovered several safety violations. Burke told Ewing there were rumors that “you blew the Darling Company into OSHA, and that you won’t go back to work for them any more.”

Ewing testified that he immediately called Roy Shafer, Darling’s vice president, who bluntly confirmed the rumor. The following day, Ewing visited the OSHA office, and was advised that their file contained no letter bearing his name. With that information, Ewing and Burke then met with Shafer who, both testified, stated he had narrowed to three the list of possible informants in the OSHA incident, and that in any case, he did not want Ewing at the company again. After Burke left the meeting, Ewing asked Shafer and Radel (who had arrived in the interim) whether there was anything wrong with his workmanship or attitude. He was told there was not.

Ewing immediately communicated with an attorney, James Kogler, to solicit aid in getting his job restored. Kogler wrote to Herbert Darling and later spoke by telephone with his attorney, Frederick Turner. Kogler testified that Turner informed him “there was a question of fact about the OSHA issue,” and that he wanted verification that Ewing had not contacted OSHA, and a statement from Burke about the Shafer-Burke-Ewing conversation. Kogler obtained a letter from OSHA, notified Turner of that fact, and claims he was told that the statement from Burke was no longer necessary. Shortly thereafter, Turner wrote a letter to Kogler confirming that Ewing would be recalled to work.

Ewing was, in fact, recalled on April 27. Ewing related that on May 4 Darling visited the jobsite, and acknowledged to him that he had been treated unjustly. Darling, Ewing testified, told him that “he wanted to at least try and make amends for it by giving me back some work.” Notwithstanding Darling’s statement, however, Ewing was laid off the same day, and worked very little over the next seven months. Burke testified he called Darling once during that period, and that Darling conceded to him that Ewing had been treated badly and that he would attempt to get him back to work. Ewing filed an unfair labor practice charge with the NLRB on July 21. He was not recalled thereafter until December, and then worked until the hearing on this matter in January 1982.

Darling’s witnesses confirmed that the conversations testified to by Ewing and Burke had occurred, but denied that the *1120 OSHA incident had ever been discussed as bearing on their failure to recall Ewing. Shafer testified that there was simply not enough work to justify recalling Ewing, and that he had heard reports of problems with Ewing’s work from his supervisors. Shafer also stated that he had known all along that, because the OSHA inspection was a general one involving the entire job-site, it could not have been triggered by a complaint about any particular job-related condition. Radel testified that Ewing had less seniority on Darling jobs than any of the other men on his team, which accounted for his seemingly disparate recall record. Radel acknowledged, however, that both Darling and Shafer at different times asked him whether he knew of any connection between Ewing and OSHA. Radel also stated that Ewing was the only man hired directly from the union hall for the rapid transit project .after December 1980, and that Radel called for him “on Mr. Darling’s request or direct orders.” Darling denied he had ever admitted that Ewing had been treated discriminatorily as a result of the alleged OSHA incident, but did concede that he “chewed Roy [Shafer] out good for any accusation at all, because it’s a violation, and it’s not our company policy, and ... you don’t do things like that.”

After a two-day hearing, and upon consideration of the full record and “observation of the witnesses and their demeanor,” the AU held that Ewing had been discriminated against for protected concerted activity in violation of § 8(a)(1) of the National Labor Relations Act (the “Act”), 29 U.S.C. § 158(a)(1). He specifically credited the testimony of Ewing and Burke, and discredited that of Shafer. He ordered Darling to cease and desist from retaliating against employees for the actual or apparent exercise of protected rights, and to make Ewing whole for the loss of pay he suffered by reason of the company’s discriminatory conduct.

The Board, however, reversed the AU’s order by a two-to-one vote and dismissed the complaint for failure to establish a labor law violation. The Board majority emphasized that Darling officials knew the OSHA inspection was of a routine nature, and was not in response to a specific complaint. In addition, the majority argued that Ewing’s sporadic work history was explained by the fact that he was the only member of his crew without a recall preference. Finally, they noted that the AU’s credibility resolutions were “not based specifically on demeanor, but rather are premised on an analysis of the facts and the logical inferencés to be drawn therefrom, and we are therefore as able to decide issues of credibility as he.” The dissenting member took issue with the Board decision, finding that it was based on evidentiary speculation rather than upon the credibility determinations on the record, and thereby “flout[ed] the time-honored role of the administrative law judge.” Ewing now petitions this court, pursuant to § 10(f) of the Act, 29 U.S.C. § 160(f), to set aside the Board’s order and reinstate the order of the AU.

As our recital of the facts makes clear, this case turns upon credibility. Normally, a reviewing court will not reject a decision of the Board unless, after canvassing the entire record, it is not supported by substantial evidence. Universal Camera Corp. v. NLRB, 340 U.S. 474, 71 S.Ct. 456, 95 L.Ed. 456 (1951). The AU’s findings, however, “are a part of the record that must be considered by the reviewing court in assessing the substantiality of the support for the Board’s findings.” NLRB v. Donald E. Hernly, Inc., 613 F.2d 457, 462 (2d Cir.1980). When the Board overturns the determinations made by an AU, therefore, its own findings “must be stronger than would be required in cases where the [AU’s] findings are accepted.” NLRB v. Interboro Contractors, Inc., 388 F.2d 495, 499 (2d Cir.1967). The significance attached to the AU’s conclusions, in relation to the Board’s, “depends largely on the importance of credibility in the particular case.” Universal Camera Corp. v. NLRB, supra, 340 U.S. at 496, 71 S.Ct. at 469. *1121 Where credibility is of central importance, and “the Board second-guesses the Examiner and gives credence to testimony which he has found ... to be inherently untrustworthy, the substantiality of that evidence is tenuous at best.” Ward v. NLRB, 462 F.2d 8, 12 (5th Cir.1972).

We find the evidence supporting the Board’s version of events far more meager than that which prompted the AU’s findings. The only witness not associated with either party, Burke, testified in support of Ewing’s statement of the case. The evidence established that Burke acted as a middleman, who matched employer needs with employee availability, and who was intent upon maintaining good relations with both. He stated that he was instructed in no uncertain terms that the Darling firm was no longer interested in Ewing’s services. 1 He also confirmed all material elements of Ewing’s allegations about Shafer’s statements with regard to the OSHA connection, and Darling’s expressions of concerns about Ewing’s mistreatment.

Moreover, of the company witnesses, only Shafer’s testimony was completely inconsistent with Ewing’s. While Darling disavowed any admission that Ewing had been treated discriminatorily as a result of the OSHA rumor, he did concede having reprimanded Shafer because the story was circulating. In addition, Radel testified that, sometime in 1981, Darling ordered him to call Burke and request that Ewing be sent out, the only time such an order was ever given. We find it difficult to explain such an extraordinary act on Darling’s part absent exceptional circumstances to warrant it.

Assuming that Ewing’s allegations are entirely fabricated, the company has failed to offer an alternative credible explanation for refusing to recall him. Early in the hearing before the AU, the parties stipulated that the quality of Ewing’s work played “no part in his being laid off, or not being recalled to work .... ” In view of this stipulation, the Board was compelled to disregard Shafer’s later insinuations that company supervisors had found Ewing a difficult employee. Nor can the failure to recall Ewing be explained by the fact that he was the only member of his piledriving crew without a recall preference. Burke testified that only two preferential rights existed under the Union’s collective bargaining agreement. A piledriver working on a job cannot be laid off and replaced by another, and a team which is laid off has a right to return to a job when work is resumed. No other seniority provisions existed, according to the statement of the Union business agent. Clearly, Ewing’s right to be recalled to the rapid transit worksite with his team was violated by Darling. 2

When piledrivers are not available for a particular project, the company is allowed to substitute “permit men” — that is, members of the Carpenter’s Union — to take their place. The record establishes, however, that the company hired permit men on three occasions while Ewing remained on layoff. The Board admits it cannot account for this fact, and suggests in a footnote that it “demonstrates only that, for one reason or another, established procedures were ignored ____” Contrary to this unsatisfactory treatment of the issue, the hiring of permit men lends strong cir *1122 cumstantial support to the claim that the company was determined not to recall Ewing regardless of its need for additional piledrivers.

Finally, the Board argues that because Shafer knew the OSHA inspection was routine, he could not have suspected Ewing of filing a complaint and therefore had no motive to discriminate against him. The strength of this argument depends precisely upon the reliability of Shafer’s testimony, which was specifically discredited by the AU. We are not persuaded that the Board may ignore the AU’s credibility determinations, or may disregard what the General Counsel describes as the AU’s “boilerplate recitation” that his findings were based on observation and demeanor of the witnesses. The Board, we think, makes an overly fine and, in this case at least, untenable distinction between credibility determinations based upon demeanor evidence and those “premised on an analysis of the facts and the logical inferences to be drawn therefrom____” If that differentiation does exist, there is no indication here that the AU’s judgments were of the latter variety.

We therefore find that the Board’s reversal of the AU’s findings that the failure to recall Ewing from layoff was motivated by the belief that he had contacted OSHA is not supported by substantial evidence. That does not, however, end the matter. The AU’s decision was properly premised on the view that the filing of a complaint by a single individual could be deemed “concerted activity” which is protected under § 7 of the Act, 29 U.S.C. § 157. The Board had approved this approach in Alleluia Cushion Co., 221 NLRB 999 (1975), which held that concertedness would be deemed to exist when an individual action reflected the collective concern of other employees in the workplace. Since its decision in the instant case, however, the Board has overturned Alleluia Cushion in Meyers Industries, 268 NLRB No. 73 (January 6, 1984). The question now before us is whether Meyers Industries should govern this case, and should be retroactively applied to events which occurred more than two years before it was decided. Because that issue was not squarely presented or briefed by the parties, we must remand to the Board for a determination of the applicability of Meyers Industries. If the Board concludes that its recent decision should not govern here, we direct it to reinstate the decision and order of the AU. Alternatively, if Meyers Industries does control, the Board is instructed to reevaluate the record in this case in light of its newly-adopted principle.

Accordingly, we remand to the Board for the foregoing determination.

. Burke did testify, in answer to meticulous questioning by Darling’s attorney, that no one from the company instructed him not to send Ewing out in the months of January or February 1981. The record establishes, however, that it was in December 1980 that Ewing’s entire team was recalled while he was not. Moreover, it was apparently in March 1981 that Ewing approached Burke about the company’s failure to recall him, resulting in the Shafer-Burke-Ewing conversation. Because of the restricted time frame applied to Burke’s testimony by Darling’s counsel, therefore, it seems to have little if any probative value.

. The General Counsel elevates to the level of a mandatory procedure what appears, instead, merely a company custom that “regular" or “steady Eddie” employees were requested by name, while nonregular employees like Ewing were sent out at Burke’s discretion. Moreover, the General Counsel’s assertion that the company never requested nonregular piledrivers by name is flatly contradicted by Radel’s testimony-