In France, before the army became a professional activity, all 18-year males, before entering military service, participated in a 2-day preliminary programme, undergoing medical, physical and psychometric testing. For our investigation, a questionnaire regarding the subject and his parents was added. A single military region, that of Ile-de-France, was investigated in March 1990 and January 1991. The population was made up of 10 809 subjects. After removing 1863 unusable questionnaires (refusal to answer, incomplete or contradictory answers, etc.), there remained 8946 subjects. To obtain as much homogeneity as possible, only subjects born to native French parents and born themselves in France, Western Europe, Canada or the United-States were included. Only records with no missing data regarding parents' ages were retained. 6564 subjects then remained. The majority of the population studied were 18- and 21-year olds (the latter being conscripts temporally exempted because of their studies). There were also two extreme limits: one consisting of volunteers who were only 16-years old and the other of late student subjects, whose maximum age was 30.

This Army investigation comprised a series of 187 items, divided in three major types of tests, based on a minimum level of education, i.e. knowing how to read and count. These tests, validated against the Wechsler scales, had been used by the French army since 1981 [23]. They analysed general aptitude for reasoning, so-called “practical intelligence” and the aptitude to use verbal symbols and numbers correctly. The answers were provided as “multiple choice” and tests were marked automatically by machine. These three types of tests have been described previously [10].

Raw scores obtained in each test were combined into a single general mark (GM) with a maximum of 20, referred to as “general intelligence level”. Military experts observed that a GM ⩾ 11 authorized access to any military occupation, while a GM < 11 more or less seriously reduced this possibility.

In our earlier investigation [10], we were restricted by military conditions to subjects' date of birth and parents' ages (and the latter were not normally required). In the present work, on the contrary, numerous questions were authorized. We could therefore employ multivariate analysis and, so, make this investigation more secure than the first. Factors studied were:

• – subject's and parents' places of birth;
• – subject's date of birth, height and weight; however, two of these factors were not maintained: (a) subject's age, because this variable was closely related to academic standard (p<.0001) and to GM (p<.0001); (b) weight, because this variable was closely related to height (p<.0001) which, being known to reflect the socio-economic status [24–26], was chosen as socio-economic indicator;
• – number of brother(s) and sister(s) (6 classes);
• – birth order (5 classes);
• – subject's and parents' current tobacco consumption before and at time of interview (at least 5 cigarettes or 2 pipes a day) (answer: yes or no); no question about pregnancy period;
• – subject's and parents' current alcohol consumption before and at time of interview (at least 1 litre of wine or equivalent a day) (yes or no); no question about pregnancy period;
• – subject's and parents' academic standard reflecting cultural status and comprising 4 classes: non-schooled or primary standard; secondary or technical school, 1st cycle; secondary or technical school, 2nd cycle; A level and more; parents' academic standards, though closely related (p<.0001), were, nevertheless, separately studied because of a possible difference between mother's and father's influences;
• – parents' socio-economic status (10 classes); this factor was not analysed because the ten classes proposed in our questionnaire proved insufficiently descriptive of actual occupations;
• – parents' dates of birth and ages at the date of the investigation; this enabled us to verify the consistency of answers and calculate parents' ages at the time of the subject's birth (father: 7 classes; mother: 6 classes). Though closely related (p<.0001), father's and mother's ages at this time were also separately studied, because of a possible difference between father's and mother's influences.

The study received the authorization of the “Service de Santé des Armées”. The investigation was anonymous.

The two GM populations were compared in the various situations presented in the questionnaire. Quantitative results expressed as means were compared using Student's t test. Qualitative results were compared using Chi-2 test. Multivariate logistic regression (SAS program, SAS Institute, version 8.2) was used to assess the independent role of selected factors and to search for possible effect modifications between some variables.

The mean GM ± sd was 12.6±5.6 ( min.: 1.0, max.: 20); the mean subjects' age was 19.8±1.7 (min.: 16.0, max.: 30.0); the mean father's age at the subjects' birth was 28.5±6.1 (min.: 15.0, max.: 62.0); the mean mother's age at the subjects' birth was 26.0±5.5 (min.: 14.0, max.: 45.0). The distribution of GMs is shown in Fig. 1. Since this distribution was not normal, we chose to split GM into two groups according to the gradation established by military experts: ⩾11 (n1=4257, 64.9%), <11 (n2=2307, 35.1%).

We compared the two groups according to the different variables of the questionnaire. Except for “subject's alcohol consumption” in which these groups did not differ, they were significantly different for the other variables. In these, the GM < 11 group was respectively related to low height, high number of siblings and junior birth order, subject's and parents' tobacco consumption, parents' alcohol consumption, subject's and parents' low academic standard, parents' youth/aging at conception.

Only variables with significant results in the univariate analysis were retained in the multivariate analysis. Nevertheless, subject's academic standard was excluded because the correlation between GMs and academic levels was so strong that it could be considered that these two parameters measured the same thing – mental aptitudes – by two different means. Through multivariate analysis, Table 2 showed that subject's low height, parents' low academic standard, junior birth order, subject's and parents' tobacco consumption, father's and mother's youth at conception, taking those aged below 20, were significantly related to GM < 11. However, such a relationship did not appear with the parents' aging.

In accordance with the Turkheimer and al.'s results [21], the parents' cultural and socio-economic status, measured in our study by the parents' academic standard and the subject's height, were found to be the most important factors for having GM ⩾11 or < 11. The role of the father's age was overall not significant (p=.12), even if the OR associated with the 15–19 year class was significantly different from one compared to the 25–29 year class as the reference. No effect modification between father's age and these variables was found.

The probability for the subject to have a GM < 11 was inversely related to:

• (a) the parents' educational level. In children, some studies showed this negative correlation concerned the mother [18], and others concerned the father [19], but via a related factor – the father's social class. Indeed, if the parents' academic standard can directly influence the progeny's cultural environment, their socio-economic status, part also of his environment, plays an analogous and important role [20,21]. However that may be, the academic achievement and higher social classes seem often associated [20,21].
• (b) the subject's height. This negative correlation has previously been described [27]. The subject's height, reflecting the socio-economic status, could involve an early undernutrition of children in low socio-economic classes [18–20,26–28] – undernutrition of which deleterious effects are known in the rat [29]. However, genetic [24,26] or psychological [30] factors could interact with socio-economic ones.

On the other hand, the probability for the subject to have a GM < 11 was positively related to the birth order and that as soon as the 2nd child. These results were in harmony with previous studies [18,22]. Nevertheless, according to some authors [31,32], only a large number of siblings was found to be associated with a decrease of child quality. Our analysis, on the contrary (Table 2), effaced the significant influence of “number of brother(s) and sister(s)” factor (Table 1) and showed that only “birth order” revealed itself as significant. Since GM < 11 did not seem related to the father's or mother's aging at conception, family changes arising with time and successive births, such as the parental availability and/or the children's behavior, could be involved.

Subject's tobacco consumption. The Odds Ratio for having GM < 11 was maximum when the subject himself – an adult – smoked (Table 2). This result could correspond to the depressive effects of high doses of nicotine on the central nervous system [33]. However, to comment precisely on tobacco's mental effects in the adult we lacked the important information, whether subjects were born of smoker or non-smoker mothers during pregnancy – the first situation being long known to be harmful for progeny's IQ [34].

Parents' tobacco consumption. Apart from prenatal exposure to tobacco is the case of passive consumption. Such a relationship was previously described [14] even at extremely low levels of exposure [15]. In our study, the relationship tobacco / GM < 11 was approximately the same in the case of the father's or the mother's tobacco consumption. Aside from a possible harmfulness of nicotine in children, this effect could be related to an oxygen deprivation [35].

Finally, our results showed that a low socio-economic and cultural level and a junior birth order were more harmful for the subject's mental aptitudes than the subject's or the parents' smoking.

Subject's and parents' alcohol consumption. The univariate analysis (Table 1) did not show a significant influence of alcohol in the subject himself, probably because the duration of the toxic impregnation was too short. On the other hand, parental alcoholism had a significant effect, more important with regard to the father than to the mother. Given that the subjects themselves were not directly affected by alcohol, it seemed possible that the degradation of family socio-economic and cultural factors due to alcohol was involved and that paternal references were more deciding, for boys, than maternal ones. Several previous results agree with such parental influence [16,17] and with the importance of the paternal one on boys [36,37].

The multivariate analysis effaced the significant influence of parental alcoholism on GM value, though the father's influence was close to significance (Table 2). Consequently, we had to consider that the possible harmful influence, on progeny, of family degradation related to the parents' alcoholism, may not be constant.

Studies relating to this point are sometimes conflicting. Some show that children born of older parents have lower IQs than those born of younger ones [8]. Relative to the mother's age, a significant correlation between youngest mothers and the poorest scores of male progeny has been found [38]. Our own works have shown, in animals [9,11] and man [10] that both father's youth and aging led to a decrease of success to psychometric tests. According to Drews et al. [18], older mothers are more likely than younger mothers to have a child with both mental retardation and another neurological condition. According to Malaspina et al. [12], young (<20 years) and old (>40 y maternal) or (>50 y paternal) ages were associated with significantly lower intellectual abilities in the offspring, which confirms our previous results [9–11]. The univariate analysis presented here (Table 1) globally found again these results regarding the father's age [10] with a similar association before age 25 (unadjusted Odds Ratio 20–24 = 1.43 [1.26–1.63]) and after 44 (unadj. OR 45–50 = 1.560 [1.003–2.427]). Moreover, it showed an analogous pattern with respect to the mother's age, which led to this association before 25 (unadj. OR 20–24 = 1.45 [1.28–1.65]) and after 39 (unadj. OR 40–50 = 1.48 [1.02–2.14]). However, the multivariate analysis (Table 2) showed that GM < 11 was only related to the youngest fathers and mothers (both < 20): father adjusted OR 15–19 = 1.62 [1.04–2.52]; mother adj. OR 14–19 = 1.82 [1.38–2.40], which would confirm, for the mother, the Zybert et al. results [38] and, in part, those of Malaspina et al. [12] (concerning the father's age, the lack of effect modification between this factor and the father's academic standard reinforced the result regarding the influence of the father's youth). On the contrary, parents' aging was not related to this GM decrease. On a related topic, there was an apparent discrepancy between (a) the association relating GM to birth order and (b) the absence of such an association between GM and maternal aging. To explain this inconsistency, we could envisage that the majority of the 1863 unworkable questionnaires (see “subjects”) corresponded to aging mothers with genetic defects and that their elimination led to a bias in sampling. Questionnaires being unworkable, particularly regarding parental ages, it was impossible to answer this question. Nevertheless, given that, in the population studied, less than 9% (that is to say a minority of the mothers) were 35–50 year old, one could suppose that proportions were the same in the eliminated population. Moreover, it seems difficult to associate systematically consequences of maternal aging with those of birth order: indeed, the latter could play a role independently of the maternal aging through, among other things, modifications of siblings' relationships and/or decrease mother's availability. As for the parental, and particularly the father's youth, given that, from Table 2, its consequences were independent from studied environmental factors, it was possible to hypothesize genetic ones to explain the GM < 11 phenomenon. These factors could involve male and/or female gametes: in the post-puberal period, due to the influence of epigenetic factors in the final structural arrangement of DNA in sperm and oocytes, there might be a “run-in” period in some way before maturity is reached [10,11]. Concerning paternal youth, such clinical situations are observed for other complex genetic conditions, including cardiac deficits [39], neural tube defects [40] and type I diabetes mellitus [41]. As for maternal youth, its role in the increase of T 21 is known [42,43]. However, in the case of young mothers, an analogous uterine run-in period might be also considered.

Concerning the parents' aging, the multivariate analysis showed, relatively to the reference age (Table 2, 25–29 years), that it did not increase the risk of mental defect, contrary to what was found by Malaspina et al. [12]. Concerning this discordance it is noticeable: (a) that psychometric tests used were not the same; (b) that Malaspina's results involve male and female adolescents – and differences between male and female successes vary according to nature of tests [44–46] – while ours concern only male adults. However that may be, we have seen above that subtle troubles could be associated to parental and, particularly, father aging. Consequently, how can we explain our own results?

It is known that the synaptic plasticity and learning are conditioned by genetic [47–50] and environmental [51–56] factors. Our own study showed the importance of some of the latter, such as socio-economic and cultural factors, regarding the GM level. Finally, and in spite of the obvious absence of effect – i.e. probably genetic effect – of parental aging at conception on progeny's GM level, we have to wonder whether genetic and environmental factors could not be both independent and complementary.

From all these elements, it is possible to hypothesize a synthesis combining genetic and environmental factors through the notion of “synaptic potential”. For example, if an individual is born with the genetic possibility to develop 6 synapses per neuronal surface unity, he will develop them if the environment is favourable but, if it is not, he may only develop 3 or 4. He will in that case perform less well than another individual born with a lower potential of 5 but who will develop them because his environment is favourable. This conception seems be in agreement with Turkheimer and al.'s observations [21]. Thus, if the gametes' genetic changes related to parental aging, those in particular concerning spermatozoa [3,9,10,12], would decrease the progeny's synaptic potential, an important environmental stimulation of the latter could compensate for this decrease. Such a stimulation could be due to an increase of socio-economic and cultural level in older parents (particularly regarding the mother since there was no effect modification between the father's age and his academic status) either through an academic course or, with time, an improvement via self-education. In this way, the absence of risk in progeny that we observed would be explained and that all the more because the homogenization we carried out in our population (see Section 2) perhaps increased the probability to reveal the better socio-cultural situations. Consequently, it would be interesting to verify that such a socio-economic and cultural increase did not concern Malaspina et al.'s aging parental population [12], possibly more heterogeneous than ours. From the general perspective of synaptic potential, let us notice that a greater risk of poor results in children from young couples could – adding itself to the run-in genetic period – result from a lower socio-economic and cultural level in these couples. Finally, genetic factors would operate on the development of synapses not at all in an exact determination of their number, but on the more or less large morphological and physiological capacity to develop them, quantitatively, under the influence of environment. This genetic capacity would vary according to individuals. In this way, it would originate an individual synaptic potential which might change with parental age, sex and cerebral areas.