On November 21, 1977, in a wooded residential area of Plainville, Connecticut, Plainville police officer Robert M. Holcomb was found lying on the ground with four bullet wounds in his body. Less than an hour later he was pronounced dead. On Febru*418ary 28, 1978, a Hartford county grand jury returned a two count indictment charging the defendant with capital felony for the intentional murder of a police officer engaged in the performance of his duties, in violation of General Statutes §§ 53a-54a (a) and (c)1 and 53a-54b,2 and with felony murder for causing the death while in the course of and in flight from a burglary or *419attempted burglary, in violation of General Statutes §§ 53a-54a and 53a-54c.3 After a jury trial, the defendant was found guilty on both counts. The trial court originally imposed consecutive sentences of not less than twenty-five years to life on each count for a total effective sentence of fifty years to life,4 but upon the defendant’s motion to correct the illegal sentence the court modified its judgment so that the defendant received a net effective sentence of twenty-five years to life.5
*420The defendant’s appeal from the judgment of conviction presents two claims of error: (1) whether the grand jury that indicted the defendant was drawn from a fair cross section of the community, as provided for by the due process clause of the fourteenth amendment to the United States constitution and (2) whether the defendant is entitled to a new trial because the trial court improperly instructed the jury that it could discuss the evidence prior to the close of all the evidence and the court’s charge.
I
The defendant’s first claim of error is that the under-representation of Hispanics on grand juries in the judicial district of Hartford county denied him his rights to due process, as provided by article first, §§ 8, 9 and 20 of the Connecticut constitution and the fourteenth amendment to the United States constitution, and that, as a result, the indictment against him should have been dismissed. We do not agree.
The essence of a due process challenge to the grand jury is that the array was not representative of a fair cross section of the community. Peters v. Kiff, 407 U.S. 493, 500-502, 92 S. Ct. 2163, 33 L. Ed. 2d 83 (1972) (plurality opinion). It is analogous to a sixth amendment claim that a petit jury was not “impartial” because it was not drawn from a fair cross section of the community. Duren v.Missouri, 439 U.S. 357, 99 S. Ct. 664, 58 L. Ed. 2d 579 (1979); Taylor v. Louisiana, 419 U.S. 522, 95 S. Ct. 692, 42 L. Ed. 2d 690 (1975). Accordingly, in its analysis, the trial court was guided by Duren v. Missouri, supra, which enunciated a test for analyzing a fair cross section challenge.
The defendant argues and the state agrees that the trial court’s analysis should have been governed by Castaneda v. Partida, 430 U.S. 482, 97 S. Ct. 1272, 51 L. Ed. 2d 498 (1977), because Duren was a sixth amendment fair cross section challenge to a petit jury, while Castaneda was a fourteenth amendment challenge *421to a grand jury. We do not agree. Because the fourteenth amendment due process claim is analogous to a sixth amendment claim, we can rely on sixth amendment cases. Cf. People v. Guzman, 60 N.Y.2d 403, 409-10 n.3, 457 N.E.2d 1143, 469 N.Y.S.2d 916 (1983). Moreover, Castaneda was solely an equal protection challenge to a grand jury and hence the test it enunciated reflects equal protection jurisprudence. There are three components to a prima facie equal protection violation in the grand jury context: (1) underrepresentation of a recognizable group; (2) substantial under-representation over a significant period of time; and (3) a selection procedure susceptible to abuse or not racially neutral. Castaneda v. Partida, supra, 494. Although the equal protection test is similar to the cross section test, the critical difference is that in an equal protection claim the defendant must prove discriminatory purpose. Once the defendant has established this prima facie case, the burden shifts to the state to rebut the presumption of discriminatory purpose. Id., 495. In contrast, in a fair cross section claim, the defendant need not prove intent. “[Systematic disproportion itself demonstrates an infringement of the defendant’s interest in a jury chosen from a fair community cross section. The only remaining question is whether there is adequate justification for this infringement.” Duren v. Missouri, supra, 368 n.26; id., 371 (Rehnquist, J., dissenting); United States v. Perez-Hernandez, 672 F.2d 1380, 1384 n.5 (11th Cir. 1982); Villafane v. Manson, 504 F. Sup. 78, 82 n.6 (D. Conn.), aff d without opinion, 639 F.2d 770 (2d Cir. 1981). Accordingly, we turn to Duren v. Missouri.
“In order to establish a prima facie violation of the fair-cross-section requirement, the defendant must show (1) that the group alleged to be excluded is a ‘distinctive’ group in the community; (2) that the representation of this group in venires from which juries are selected is not fair and reasonable in relation to the number of such persons in the community; and (3) that this underrepresentation is due to systematic exclusion *422of the group in the jury-selection process.” Duren v. Missouri, supra, 364. Once the defendant has established this prima facie case, the burden then shifts to the state to prove that the selection system resulting in a nonrepresentative array furthers a significant state interest. Id., 367.
The trial court found that the defendant had established a prima facie case. It first determined that Hispanics are a cognizable group in Hartford county and that the defendant satisfied the first prong of Duren. With respect to the second prong, the trial court found that from January 1, 1970, until February 27, 1978 (this latter date representing the defendant’s indictment), 972 grand jurors served in Hartford county of whom 17 or 1.75 percent were Hispanic. For the period of 1970-1977, the mean proportion of Hispanics in the general population was between 3.10 percent and 4.69 percent.6 The court concluded that a comparison between the highest mean population figure, 4.69 percent, and the percentage of Hispanic grand jurors, 1.75, made it “clear that Hispanics were not represented on the grand jury in proportion to their number in the county population.”
Turning to the third prong of Duren, the court found that the major cause of Hispanic underrepresentation is the requirement of General Statutes § 54-45 that grand jurors be electors. It noted that the general population rate of voter registration is 3.5 times the registration rate for Hispanics. The court applied this factor to the highest mean Hispanic population figure of 4.69 percent, and found that, for the period of 1970-1977, Hispanics represented only 1.34 percent of the general voting population of Hartford county. It then compared *423this figure with the percentage of Hispanic grand jurors and concluded “that Hispanic voters were overrepresented on the grand jury during that period by about 30%.” (Emphasis added.)
The court went on to determine that even though the use of voter lists as a source for selection resulted in underrepresentation of Hispanics, such a requirement was justified “because registration as a voter is at least some indication of an interest, ability, and willingness to participate in the governmental process of which the function of the grand jury is a part.” Finally, the court noted that even if the voter registration requirement were eliminated, the underrepresentation of Hispanics on grand juries would probably continue because of “their deficiency in the use of the English language in which grand jury proceedings are conducted” and “other factors having a special impact upon that group.” It therefore determined that the state had rebutted the defendant’s prima facie case and it denied the defendant’s motion to dismiss the indictment.
On appeal, the defendant claims that the trial court erred in holding that the state had rebutted the defendant’s prima facie case of unconstitutional discrimination. He contends that there was insufficient evidence in the record to support the court’s conclusions that Hispanics had a low level of voter registration and English language ability and that these factors in fact caused the underrepresentation of Hispanic grand jurors. The state responds that we need not reach this issue because the defendant failed to make out a prima facie case. Though we do not agree with the state’s reasoning, we do conclude that the defendant failed to prove that there was substantial underrepresentation of Hispanics.
*424COGNIZABILITY
The state concedes that Hispanics7 comprise a distinctive group in Hartford county. We agree that the defendant satisfied the first prong of Duren.
Proof of Underrepresentation
In finding that the defendant established proof of underrepresentation, the trial court compared the percentage of Hispanics in the general population of Hartford county (between 3.10 percent and 4.69 percent) with the percentage of Hispanics who served on Hartford county grand juries (1.75 percent). The state contends that the court should have compared the percentage of Hispanic grand jurors with the percentage of those Hispanics in the population eligible to serve as grand jurors, in this case Hispanics registered to vote. Under the state’s approach, the defendant could not satisfy this second prong because, as the trial court found by comparing 1.75 percent with 1.34 percent, Hispanics were overrepresented on grand juries during the relevant period.
We recognize that there is considerable controversy about this issue. In Castaneda v. Partida, supra, 494, an equal protection case, the majority compared “the proportion of the group in the total population to the proportion called to serve as grand jurors, over a significant period of time.” In a dissent joined by two other justices, Chief Justice Burger termed this a flaw and argued that the relevant statistical universe was eligi*425ble population. Id., 504-505; see Alexander v. Louisiana, 405 U.S. 625, 630, 92 S. Ct. 1221, 31 L. Ed. 2d 536 (1972). Some lower courts have distinguished Castaneda and relied upon eligible population statistics. See, e.g., United States v. Brummitt, 665 F.2d 521, 529 (5th Cir. 1981), cert. denied, 456 U.S. 977, 102 S. Ct. 2244, 72 L. Ed. 2d 852 (1982); United States ex rel. Barksdale v. Blackburn, 639 F.2d 1115, 1123 (5th Cir.), cert. denied, 454 U.S. 1056, 102 S. Ct. 603, 70 L. Ed. 2d 593 (1981). But see Bryant v. Wainwright, 686 F.2d 1373, 1377 (11th Cir. 1982); see also Villafane v. Manson, 504 F. Sup. 78, 86-87 n.13 (D. Conn. 1980), aff'd without opinion, 639 F.2d 770 (2d Cir. 1981) (assuming without deciding that the relevant population is eligible population).
This controversy does not affect our analysis, however, because as we previously indicated, this is a fair cross section claim, not an equal protection claim. In Duren, a case with only a single dissenter, the court stated (p. 365 n.23), “the fair-cross-section requirement involves a comparison of the makeup of jury venires or other sources from which jurors are drawn with the makeup of the community, not voter registration lists.” (Emphasis in original.) Hence the trial court made the proper comparison.8
Where we depart from the trial court is in its finding that the disparity between the percentage of Hispanics who served as grand jurors and the percentage of Hispanics in the population is constitutionally significant. “The jury array need not mirror the sociological composition of the community; Swain v. Ala*426bama, 380 U.S. 202, 208, 85 S. Ct. 824, 13 L. Ed. 2d 759 (1965); or be mathematically proportionate to the percentage of the group in the community. Duren v. Missouri, supra. Substantial underrepresentation is required to make out a prima facie case.” State v. Haskins, 188 Conn. 432, 440, 450 A.2d 828 (1982). The trial court did not state which statistical method it utilized in concluding that the defendant proved substantial underrepresentation. Had the trial court used the method most appropriate to examine the defendant’s claim, it could not have concluded that the defendant had proved substantial underrepresentation. As a result we must enter the statistical morass that has developed in this area.
The application of the science of statistics to legal issues is becoming increasingly common because it is helpful in measuring and evaluating disparities. This tool is not without its hazards, however. When applied to measure concepts such as intent and fairness, it is necessarily imprecise. This is further complicated by the fact that there are a number of statistical methods that have developed and, depending on which one is used, a different result is produced. Moreover, there is no consensus among the courts, commentators, or statisticians on which method is most reliable. See Baldus & Cole, Statistical Proof of Discrimination § 5.1, n.9 (1980). Compare United States v. Haley, 521 F. Sup. 290, 292 (N.D. Ga. 1981) (prefers absolute disparity method), with Quadra v. Superior Court, 403 F. Sup. 486, 495 n.9 (N.D. Cal. 1975) (prefers comparative disparity method). Indeed, the United States Supreme Court has left this decision to the courts. See Alexander v. Louisiana, 405 U.S. 625, 630, 92 S. Ct. 1221, 31 L. Ed. 2d 536 (1972).
Our research has revealed that the choice of a statistical method depends on the facts and circumstances of each case. An important consideration is whether the *427claim is an equal protection challenge, where the issue is whether the discrimination was purposeful or the result of chance, or whether the claim is a fair cross section challenge, where the issue is whether the under-representation is unfair.9 Another important factor is the size of the population at issue. Ultimately, however, the decision is not one of numbers but rather a subjective determination of whether the disparity is constitutionally significant. Cf. United States v. Test, 550 F.2d 577, 589 (10th Cir. 1976).
There are essentially four10 statistical models that have been developed to analyze grand jury discrimination claims. The one the defendant suggests we employ is the Statistical Decision Theory (SDT). First expounded in 1966; Finkelstein, “The Application of Statistical Decision Theory to the Jury Discrimination Cases,” 80 Harv. L. Rev. 338, 349-53 (1966); SDT has gradually become the favored method of the United States Supreme Court in analyzing equal protection challenges to grand jury arrays. Castaneda v. Partida, supra, 496 n.17.11 See also Villafane v. Manson, supra, 85-86, and State v. Villafane, 164 Conn. 637, 638, 325 A.2d 251 (1973). The import of this method is that it is a highly sophisticated and precise indicator of whether the disparity between the group’s representation in the popu*428lation12 and its representation among grand jurors is the result of a neutral and random selection process. In the context of a fair cross section claim, however, where the focus is not on intent but rather on whether the array is reasonably representative of the community, SDT is inapposite.
The other methods are relevant to a fair cross section claim. The first and most widely used is the absolute difference test. This method measures the difference between the percentage of the cognizable class in the population and the percentage of that group represented in the venire. United States v. Haley, 521 F. Sup. 290, 292 (N.D. Ga. 1981). In our case, using the highest population figure, the result is an absolute difference of 2.94 percent (4.69 percent minus 1.75 percent), which can hardly be deemed constitutionally significant. See Swain v. Alabama, 380 U.S. 202, 208-209, 85 S. Ct. 824, 13 L. Ed. 2d 759 (1965) (absolute difference of approximately 10 percent not substantial), and cases cited in State v. Haskins, supra, 441. We cannot rely on this method, however, for though it is useful to measure the underrepresentation of a group that represents a large proportion of the population; see State v. Haskins, supra; it is inadequate when the percentage of persons in the group is small in relation to the entire population. The result obtained is a distortion of reality.13
*429The second method is known as the ratio method, or comparative disparity method. This is calculated by the following formula:
Proportion of the population that is in the specified category
-
Proportion of the pool that is in the specified category
Proportion of the population that is in the specified category
National Jury Project, Jurywork-Systematic Techniques § 5.05 [2] [c] [ii] (2d Ed. 1983). If we apply this formula to the figures in our case the result is a comparative disparity of approximately 63 percent (4,64-1.75/4.69). Although upon first glance this appears to reflect a substantial underrepresentation of Hispanics, this figure is misleading. “It is necessary to realize that the comparative disparity calculation tends to magnify the size of the disparity as the relevant group’s percentage of the population decreases.” United States v. Haley, supra, 292. “The defendant characterizes the deviation in comparative terms and says that it exceeds 80%. While such a characterization may be proper where blacks constitute a significant proportion of the population, it is ordinarily inappropriate where a very small proportion of the population is black. A comparative characterization in such circumstances distorts reality.” United States v. Musto, 540 F. Sup. 318, 355 (D.N. J. 1982), aff’d sub nom. United States v. Aimone, 715 F.2d 822 (3d Cir. 1983), quoting United States v. Whitley, 491 F.2d 1248, 1249 (8th Cir.), cert. denied, 416 U.S. 990, 94 S. Ct. 2399, 40 L. Ed. 2d 769 (1974); see also Villafane v. Manson, supra, 84. Because the Hispanic population in this case is small, we cannot rely on this method.14 Cf. National Jury Project, supra, § 5.05 [2] [c] [ii] n.109.
*430The third method is called the substantial impact test. Its focus is not on numbers and percentages but rather on whether the underrepresentation substantially affects the composition of the grand jury. United States v. Test, 550 F.2d 577 (10th Cir. 1976); United States v. Goff, 509 F.2d 825 (5th Cir.), cert. denied, 423 U.S. 857, 96 S. Ct. 109, 46 L. Ed. 2d 83 (1975); United States v. Jenkins, 496 F.2d 57 (2d Cir. 1974), cert. denied, 420 U.S. 925, 95 S. Ct. 1119, 43 L. Ed. 2d 394 (1975). When this method is applied to the facts of our case, it becomes clear that the defendant cannot establish substantial impact because of the underrepresentation of Hispanics. The number of grand jurors during the relevant period was 972 (54 grand juries x 18, the number of grand jurors per panel). We will compare this to the trial court’s highest population estimate, 4.69 percent. If this array mirrored the number of Hispanics in the population, it should have included almost forty-six Hispanics (4.69 percent of 972). In fact this array included seventeen Hispanics (1.75 percent of 972). Hispanics were therefore underrepresented in this array by twenty-nine grand jurors. Put in substantial impact terms, slightly more than one additional Hispanic should have been included on every other grand jury. Common sense dictates that the impact of this underrepresentation is not substantial. Cf. United States v. Armsbury, 408 F. Sup. 1130, 1138 (D. Or. 1976).
In analyzing a due process challenge to a grand jury array our ultimate concern is to determine whether the selection system was unfair. Thus, “[one cannot] overlook the fact that in the day-to-day operation of the jury system, the criminal defendant is not indicted or convicted by the community at large, but rather by relatively small groups of 12 or [18] persons selected to represent the community on petit and grand juries. In assessing whether a given defendant’s constitutional *431or statutory rights have been violated through the operation of a jury selection process, the proper focus of inquiry must therefore be the impact of the challenged process on the grand and petit juries. This assessment must be made in light of the well-settled rule that a defendant has no right to a grand or petit jury of any given demographic composition, but only to jury panels selected from a source ‘reasonably representative’ of the community. E.g., Taylor v. Louisiana, supra, 419 U.S. at 538, 95 S. Ct. 692 [42 L. Ed. 2d 690 (1975)]; Alexander v. Louisiana, supra, 405 U.S. at 628, 92 S. Ct. 1221 [31 L. Ed. 2d 536 (1972)]. It must likewise be remembered that our jury system of necessity deals with living individuals rather than fractional percentage persons. Changes in the demographic composition of juries and jury panels can therefore only be made by the addition or deletion of one or more individuals. Both Congress and the courts have been mindful of this latter fact in concluding that only ‘gross’ or ‘marked’ disparities or ‘substantial’ departures from a ‘fair cross section’ of the community require judicial intervention.” United States v. Test, 550 F.2d 577, 590 (10th Cir. 1976).
In rejecting the defendant’s grand jury claim, the trial court concluded that the state was justified in relying on the voter registration lists as a source for selecting grand jurors. Though this practice has been routinely validated; see, e.g., 28 U.S.C. § 1861 et seq.; United States v. Brummitt, supra, 529; State v. Townsend, 167 Conn. 539, 547, 356 A.2d 125, cert. denied, 423 U.S. 846, 96 S. Ct. 84, 46 L. Ed. 2d 67 (1975); in light of our holding that the defendant failed to satisfy the second prong of Duren, the trial court need not have reached this issue.
II
The defendant’s trial, which would last approximately six weeks, commenced on February 7, 1980, before a *432jury of twelve and four alternates. On the fourth day of trial, during the morning recess, the jury sent the trial court a note asking whether it could discuss the case as the evidence came in during trial. Over the defendant’s objection, the court recalled the jury and gave the following instruction: “The answer to the question is that you may discuss the testimony which takes place here in the courtroom among yourselves, in the jury room, when you all are present, but other than that you may not discuss the testimony.
“Of course, I want to caution you particularly about visiting the site or bringing anything into the jury room which doesn’t take place here in the courtroom. In other words, all of your discussion must be confined to what occurs here in the courtroom. I assume you have not been reading the news media so I’m not really concerned about that, but that’s part of the overall picture, too.
“I also want to say that although you may discuss the testimony as it progresses, you must refrain from reaching any conclusions because obviously the trial takes a long time, and until you have heard the whole case you shouldn’t come to any conclusions, even tentative ones. Just be careful about that and try to keep an open mind until you have heard all the testimony.” On four subsequent occasions the court reminded the jury of its instruction. The defendant asserts that these instructions violated his constitutional rights by undermining the jury’s ability to remain impartial throughout the trial.15
*433In State v. Washington, 182 Conn. 419, 438 A.2d 1144 (1980), we were faced with an almost identical circumstance. In that case, in response to an inquiry by the jury and over the defendant’s objection, the trial court instructed the jury that it “may discuss the testimony and the evidence that you hear in this case in the courtroom among yourselves .... Now, that means discuss, it does not mean deliberate. It does not mean that you come to any decision, it does not mean that you come to a conclusion. It doesn’t mean that you come to a vote. It means you discuss, talk about it if you want. As it evolved. . . . The caution that you must take is that you should never deliberate, take a vote and come to a conclusion.” Id., 423. We concluded that the danger of allowing the jurors to discuss the case before all the evidence is presented and the court instructs on the law is that in the course of the discussion a juror may form and state an opinion before he has heard the countervailing evidence and may then be reluctant to be persuaded otherwise, either by the evidence or by other jurors. Id., 426-28. We held that this instruction jeopardized the defendant’s right to an impartial jury, in violation of the sixth and fourteenth amendments to the United States constitution and article first, § 8, of the Connecticut constitution.16 “[I]t is error of constitutional magnitude for the trial judge expressly to instruct the jurors that they may discuss the case among themselves prior to its submission to them . . . .” Id., 429.
*434Both parties agree that the instruction in the present case is “essentially indistinguishable” from the one we condemned in Washington, and that such an instruction is erroneous.17 The issue in dispute is the remedy for such an error.
In Washington, supra, 429, we stated: “We do not decide that any juror discussion prior to deliberation would automatically mean a new trial. . . . What we do hold is that . . . unless the state can show that such [constitutional] error is harmless beyond a reasonable doubt, the defendant is entitled to a new trial.” It is the defendant’s position that because there was so much opportunity for discussion, it is fair to assume that the jurors did engage in discussion that tainted their impartiality and therefore the state cannot meet this burden and a new trial is required.18
The state’s response is in the form of three alternatives. Its first and principal claim is that though the policy behind Washington is sound, that case should be overruled because it presumes prejudice unless the state can prove harmlessness beyond a reasonable doubt. The state claims that it should be the defendant’s burden to prove actual prejudice and it therefore requests a remand so that the defendant can prove that the instruction rendered the jurors partial. Second, the state claims that the length (three days) and nature of the deliberations indicate that the jury did not decide the case until after submission and that the evidence of the defendant’s guilt was overwhelming so that the error must be considered harmless. Lastly, the state *435requests that the case be remanded for an evidentiary hearing to provide the state with the opportunity to prove that the error was harmless beyond a reasonable doubt.
We decline the state’s invitation to overrule Washington. Though blind adherence to precedent is of no value, the necessity of certainty and continuity in the law dictates that “a court should overrule its own precedents for only the most compelling reasons. Herald Publishing Co. v. Bill, 142 Conn. 53, 62, 111 A.2d 4 (1955).” Society for Savings v. Chestnut Estates, Inc., 176 Conn. 563, 570, 409 A.2d 1020 (1979).
The state concedes that the error is of constitutional dimension. It also concedes that the weight of authority is in accord with our position that the state must bear the burden of proving harmless error. See cases cited in State v. Washington, supra, 426-27, and annot., 21 A.L.R.4th 435. It suggests, however, that we follow the reasoning of the dissent in Winebrenner v. United States, 147 F.2d 322, 329 (8th Cir.), cert. denied, 325 U.S. 863, 65 S. Ct. 1197, 89 L. Ed. 1983 (1945), and of the Maryland Appellate Court in Wilson v. State, 4 Md. App. 192, 242 A.2d 194 (1968), cert. denied, 394 U.S. 975, 89 S. Ct. 1467, 22 L. Ed. 2d 754 (1969), which requires the defendant to prove actual prejudice. These are not recent opinions and we were aware of their reasoning when we fully explored the constitutional issue and all its ramifications in Washington. Having implicitly rejected the state’s argument previously, we see no compelling reason to reconsider our position.19
We also reject the state’s second alternative. The state asks us to conclude that since the jury deliber*436ated for three days and requested the reading of certain testimony, it had remained impartial throughout the trial. Washington requires much more than speculation in order for the state to meet its burden of proof. “Here, there has been no showing that the jurors did not discuss the case. In the absence of a fair indication to the contrary, the jury is presumed to have followed the instructions of the court.” Id., 429. In addition, the state has attempted to satisfy this burden by pointing out that the evidence of the defendant’s guilt was overwhelming. This hardly addresses the issue of whether the jury that weighed that evidence was impartial.
We do not, however, agree with the defendant that reversal is required. Having stated above that the state cannot rely on speculation, we certainly will not accept the defendant’s speculation that the jurors did in fact discuss the evidence and evaluate it. We believe the state should have the opportunity it has requested to meet the burden imposed on it by Washington to prove harmlessness beyond a reasonable doubt. Accordingly, we remand the case to the trial court for an evidentiary hearing on harmless error.
A few words about the scope of this hearing: The defendant asserts that the jurors can only be asked if they did in fact discuss the evidence, with an affirmative answer resulting in a new trial. He contends that any further inquiry would violate the prohibition against probing the jurors’ mental processes. See, e.g., McNamee v. Woodbury Congregation of Jehovah’s Witnesses, 193 Conn. 15, 19, 475 A.2d 262 (1984); Williams v. Salamone, 192 Conn. 116, 121, 470 A.2d 694 (1984); *437Josephson v.Meyers, 180 Conn. 302, 310-11, 429 A.2d 877 (1980); Aillon v. State, 168 Conn. 541, 551, 363 A.2d 49 (1975); see also Practice Book § 871.20 He further points to Washington, supra, 429, wherein we stated: “Here, there has been no showing that the jurors did not discuss the case. In the absence of a fair indication to the contrary, the jury is presumed to have followed the instructions of the court.” Though we agree that the inquiry must be narrow, we do not agree that it must be as limited as the defendant suggests.
We recently stated in Williams v. Salamone, supra, 122 n.7, that “the trial court is not concerned with mental processes of the jurors, but the nature and quality of the misconduct. See Aillon v. State, 168 Conn. 541, 550-51, 363 A.2d 49 (1975). The question is whether the misconduct is of such a nature that it probably rendered the juror unfair or partial.” In determining the “nature and quality” of the “misconduct” we must be mindful that the concerns underlying Washington were not simply that the jurors may have discussed the evidence pre-submission, but that they may have taken positions on the evidence. Therefore, it would be proper to ask the jurors if they discussed the evidence. If the answer is “yes,” then it would be permissible to inquire whether the discussion was merely to refresh their memories about the evidence, in which case the erroneous instruction is harmless, or whether anyone evaluated or stated an opinion on the evidence, in which case a new trial is required. Any inquiry into the content of the opinion or the impact it had on the juror is clearly impermissible. Aillon v. State, supra, 551-52.
*438There is error on the defendant’s appeal, the case is remanded to the trial court for further proceedings consistent with this opinion. The state’s appeal is dismissed, sua sponte, for lack of jurisdiction.
In this opinion the other judges concurred.
At the time of the offense, General Statutes § 53a-54a provided: “Sec. 53a-54a. MURDER DEFINED. AFFIRMATIVE DEFENSES. EVIDENCE OF MENTAL condition, classification, (a) A person is guilty of murder when, with intent to cause the death of another person, he causes the death of such person or of a third person or causes a suicide by force, duress or deception; except that in any prosecution under this subsection, it shall be an affirmative defense that the defendant acted under the influence of extreme emotional disturbance for which there was a reasonable explanation or excuse, the reasonableness of which is to be determined from the viewpoint of a person in the defendant’s situation under the circumstances as the defendant believed them to be, provided nothing contained in this subsection shall constitute a defense to a prosecution for, or preclude a conviction of, manslaughter in the first degree or any other crime.
“(b) Evidence that the defendant suffered from a mental disease, mental defect or other mental abnormality is admissible, in a prosecution under subsection (a), on the question of whether the defendant acted with intent to cause the death of another person.
“(c) Murder is punishable as a class A felony unless it is a capital felony and the death penalty is imposed as provided by section 53a-46a.”
At the time of the offense, General Statutes § 53a-54b provided: “Sec. 53a-54b. capital felony defined. A person is guilty of a capital felony who is convicted of any of the following: (1) Murder of a member of the division of state police within the department of public safety or of any local police department, a chief inspector or inspector in the division of criminal justice, a sheriff or deputy sheriff, a constable who performs criminal law enforcement duties, a special policeman appointed under section 29-18, an official of the department of correction authorized by the commissioner of correction to make arrests in a correctional institution or facility, or of any fireman, as defined in subsection (10) of section 53a-3, while such victim was acting within the scope of his duties; (2) murder committed by a defendant who is hired to commit the same for pecuniary gain or murder committed by one who is hired by the defendant to commit the same for pecuniary gain; (3) murder committed by one who has previously been convicted of intentional murder or murder committed in the course of commission of a felony; (4) murder committed by one who was, at the time of commission of the murder, under sentence of life imprisonment; (5) murder by a kidnapper of a kidnapped person during the course of the kidnapping or before such person is able to return or be returned to safety; (6) the illegal sale, for gain, of cocaine, heroin or methadone to a person who dies *419as a direct result of the use by him of such cocaine, heroin or methadone, provided such seller was not, at the time of such sale, a drug-dependent person.”
At the time of the offense General Statutes § 53a-54c provided: “Sec. 53a-54c. felony murder. A person is guilty of murder when acting either alone or with one or more persons, he commits or attempts to commit robbery, burglary, kidnapping, arson, sexual assault in the first degree, sexual assault in the first degree with a firearm, sexual assault in the third degree, sexual assault in the third degree with a firearm, escape in the first degree, or escape in the second degree and, in the course of and in furtherance of such crime or of flight therefrom, he, or another participant, if any, causes the death of a person other than one of the participants, except that in any prosecution under this section, in which the defendant was not the only participant in the underlying crime, it shall be an affirmative defense that the defendant: (A) Did not commit the homicidal act or in any way solicit, request, command, importune, cause or aid the commission thereof; and (B) was not armed with a deadly weapon, or any dangerous instrument; and (C) had no reasonable ground to believe that any other participant was armed with such a weapon or instrument; and (D) had no reasonable ground to believe that any other participant intended to engage in conduct likely to result in death or serious physical injury.”
Prior to trial, the trial court had declared the death penalty statute unconstitutional.
The state, pursuant to General Statutes (Rev. to 1981) § 54-96, sought and received permission to appeal the correction of the sentence (Appeal No. 11166). In its brief it challenged the trial court’s ruling that the original sentence was violative of the double jeopardy clause of the fifth amendment to the federal constitution and alternatively it argued that the case should be remanded for modification of the judgment and resentencing on burglary as a lesser included offense of felony murder. It concedes that it never presented this second claim to the trial court.
Prior to and at oral argument, the state expressly withdrew the first claim of its appeal. Since that could have been the only ground for which the trial court gave permission to appeal, we have no jurisdiction to consider the remaining claim. Accordingly, we dismiss sua sponte the state’s appeal for lack of jurisdiction.
The reason that the trial court could not produce an exact figure for the proportion of Hispanics in the total population was because it had to adjust the 1970 census figure to reflect an Hispanic population that had greatly increased by 1977. To accomplish this it resorted to school enrollment figures and expert testimony presented by the defendant. This supplemental information was understandably imprecise and based on estimates.
The defendant asserts that the trial court mislabeled the cognizable group as Puerto Ricans instead of Hispanics. Though the defendant is correct that Puerto Ricans comprised 90 percent but not all of the Hispanic population in Hartford county during the relevant period, this slight discrepancy is irrelevant because in its statistical analysis the trial court explicitly utilized figures for the Hispanic population. Thus, in effect the trial court defined the cognizable group as Hispanics. We will therefore refer to the class as Hispanics.
We recognize that in State v. Haskins, 188 Conn. 432, 440, 450 A.2d 828 (1982), we erroneously utilized figures representing the eligible population. That does not change the result in that case, however, since the proper comparison would not have yielded a constitutionally significant disparity. Nor does it affect the soundness of the other propositions of law articulated in that opinion.
Our research has revealed that in the area of jury and grand jury discrimination challenges, when discussing the statistical methods many courts and commentators have ignored the different inquiries inherent in the different claims. See, e.g., United States v. Brummitt, 665 F.2d 521, 527 (5th Cir. 1981), cert. denied, 456 U.S. 977, 102 S. Ct. 2244, 72 L. Ed. 2d 852 (1982).
In Villafane v. Manson, 504 F. Sup. 78 (D. Conn. 1980), aff'd without opinion, 639 F.2d 770 (2d Cir. 1980), four methods were identified and analyzed. Although additional models exist, we similarly limit our discussion to the four most widely used methods.
This model has also been used by the court in discrimination cases outside the grand jury context. See, e.g., Hazelwood School District v. United States, 433 U.S. 299, 308-309 n.14, 97 S. Ct. 2736, 53 L. Ed. 2d 768 (1977); Washington v. Davis, 426 U.S. 229, 241, 96 S. Ct. 2040, 48 L. Ed. 2d 597 (1976).
As noted earlier, there is considerable controversy about whether the relevant population is population in the community or the eligible population. For a helpful discussion of the application of SDT, see Stone, “Grand Jury Discrimination Challenges: Defeat by Default,” 3 W. New Eng. L. Rev. 665 (1981).
“For example, an absolute disparity of 10% in a jurisdiction that is 60% black is quite different from the same disparity in a jurisdiction that is 11% black; in the latter jurisdiction, the 10% absolute disparity amounts to almost total exclusion of black people. The absolute disparity standard yields the same results in situations in which the results clearly should differ, and validates even very large underrepresentations of small and medium-sized minorities.” National Jury Project, Jurywork-Systematic Techniques § 5.05 [2] [c] [i] (2d Ed. 1983).
In addition, this method is more relevant in an equal protection case because it is probative of intent. Baldus & Cole, Statistical Proof of Discrimination §§ 5.1, 5.122 [2] (1980); cf. National Jury Project, Jury Work-Systematic Techniques § 5.05 [2] [c] [ii] (2d. Ed. 1983).
The defendant also claims that these instructions violated his statutory and constitutional rights by subjecting the jury to the outside influence of alternate jurors. The defendant raises this claim for the first time on appeal. Because of our disposition of his principal claim, it is unnecessary for us to determine whether we should consider this claim under State v. Evans, 165 Conn. 61, 327 A.2d 576 (1973).
“The sixth amendment to the United States constitution provides, inter alia: ‘In all criminal prosecutions, the accused shall enjoy the right to a speedy and public trial, by an impartial jury of the State and district wherein the crime shall have been committed . . . This guarantee has been made applicable to the states by incorporation into the fourteenth amendment. See Duncan v. Louisiana, 391 U.S. 145, 88 S. Ct. 1444, 20 L. Ed. 2d 491 (1968).
“The constitution of Connecticut, article first, § 8, provides in pertinent part: ‘In all criminal prosecutions, the accused shall have a right ... to a speedy, public trial by an impartial jury.’ ” State v. Washington, 182 Conn. 419, 425 n.5, 438 A.2d 1144 (1980).
Though Washington was decided after the defendant’s trial, there is no contention that it is not to be applied retroactively. See State v. Johnson, 188 Conn. 515, 518A, 450 A.2d 861 (1982).
At oral argument the defendant softened his position somewhat. He agreed with the “concept” of a remand to provide the state with the opportunity to prove harmlessness beyond a reasonable doubt, but asserted that in the face of the numerous opportunities to discuss, such a remand would be an exercise in futility.
The state contends that this error should be treated no differently from other juror misconduct where the defendant must prove actual prejudice. See, e.g., Smith v. Phillips, 455 U.S. 209, 102 S. Ct. 940, 71 L. Ed. 2d 78 (1982); State v. Almeda, 189 Conn. 303, 455 A.2d 1326 (1983). We disagree. A critical difference is that this “misconduct” is authorized by the trial court *436and, since jurors are presumed to follow the court’s instruction; State v. Barber, 173 Conn. 153, 156-57, 376 A.2d 1108 (1977); we must presume that they did in fact discuss the evidence. Since the line separating discussion from deliberation is a very narrow one; see State v. Washington, 182 Conn. 419, 428, 438 A.2d 1144 (1980); it is entirely reasonable to presume prejudice unless the state can disprove it.
Practice Book § 871, entitled “Impeachment of Verdict,” provides: “Upon an inquiry into the validity of a verdict, no evidence shall be received to show the effect of any statement, conduct, event or condition upon the mind of a juror nor any evidence concerning mental processes by which the verdict was determined. Subject to these limitations, a juror’s testimony or affidavit shall be received when it concerns any misconduct which by law permits a jury to be impeached.”